diff --git a/configs/_base_/datasets/decompression_test_config.py b/configs/_base_/datasets/decompression_test_config.py new file mode 100644 index 0000000000..0a52247a2e --- /dev/null +++ b/configs/_base_/datasets/decompression_test_config.py @@ -0,0 +1,66 @@ +quality = 10 +test_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='RandomJPEGCompression', + params=dict(quality=[quality, quality], color_type='color'), + bgr2rgb=True, + keys=['img']), + dict(type='PackEditInputs') +] + +classic5_data_root = 'data/Classic5' +classic5_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type='BasicImageDataset', + metainfo=dict(dataset_type='classic5', task_name='CAR'), + data_root=classic5_data_root, + data_prefix=dict(img='', gt=''), + pipeline=test_pipeline)) +classic5_evaluator = [ + dict(type='PSNR', prefix='Classic5'), + dict(type='SSIM', prefix='Classic5'), +] + +live1_data_root = 'data/LIVE1' +live1_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type='BasicImageDataset', + metainfo=dict(dataset_type='live1', task_name='CAR'), + data_root=live1_data_root, + data_prefix=dict(img='', gt=''), + pipeline=test_pipeline)) +live1_evaluator = [ + dict(type='PSNR', prefix='LIVE1'), + dict(type='SSIM', prefix='LIVE1'), +] + +# test config +test_cfg = dict(type='MultiTestLoop') +test_dataloader = [ + classic5_dataloader, + live1_dataloader, +] +test_evaluator = [ + classic5_evaluator, + live1_evaluator, +] diff --git a/configs/_base_/datasets/denoising-gaussian_color_test_config.py b/configs/_base_/datasets/denoising-gaussian_color_test_config.py index c06f0a4c3d..7950bebf35 100644 --- a/configs/_base_/datasets/denoising-gaussian_color_test_config.py +++ b/configs/_base_/datasets/denoising-gaussian_color_test_config.py @@ -17,7 +17,7 @@ params=dict( noise_type=['gaussian'], noise_prob=[1], - gaussian_sigma=[sigma * 255, sigma * 255], + gaussian_sigma=[sigma, sigma], gaussian_gray_noise_prob=0), keys=['img']), dict(type='PackEditInputs') diff --git a/configs/_base_/datasets/denoising-gaussian_gray_test_config.py b/configs/_base_/datasets/denoising-gaussian_gray_test_config.py index 326aed8a3b..774869fd07 100644 --- a/configs/_base_/datasets/denoising-gaussian_gray_test_config.py +++ b/configs/_base_/datasets/denoising-gaussian_gray_test_config.py @@ -19,7 +19,7 @@ params=dict( noise_type=['gaussian'], noise_prob=[1], - gaussian_sigma=[sigma * 255, sigma * 255], + gaussian_sigma=[sigma, sigma], gaussian_gray_noise_prob=1), keys=['img']), dict(type='PackEditInputs') diff --git a/configs/swinir/README.md b/configs/swinir/README.md new file mode 100644 index 0000000000..2ada8e12e6 --- /dev/null +++ b/configs/swinir/README.md @@ -0,0 +1,512 @@ +# SwinIR (ICCVW'2021) + +> [SwinIR: Image Restoration Using Swin Transformer](https://arxiv.org/abs/2108.10257) + +> **Task**: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + + + +## Abstract + + + +Image restoration is a long-standing low-level vision problem that aims to restore high-quality images from low-quality images (e.g., downscaled, noisy and compressed images). While state-of-the-art image restoration methods are based on convolutional neural networks, few attempts have been made with Transformers which show impressive performance on high-level vision tasks. In this paper, we propose a strong baseline model SwinIR for image restoration based on the Swin Transformer. SwinIR consists of three parts: shallow feature extraction, deep feature extraction and high-quality image reconstruction. In particular, the deep feature extraction module is composed of several residual Swin Transformer blocks (RSTB), each of which has several Swin Transformer layers together with a residual connection. We conduct experiments on three representative tasks: image super-resolution (including classical, lightweight and real-world image super-resolution), image denoising (including grayscale and color image denoising) and JPEG compression artifact reduction. Experimental results demonstrate that SwinIR outperforms state-of-the-art methods on different tasks by up to 0.14~0.45dB, while the total number of parameters can be reduced by up to 67%. + + + +
+ +
+ +## Results and models + +### **Classical Image Super-Resolution** + +Evaluated on Y channels, `scale` pixels in each border are cropped before evaluation. +The metrics are `PSNR / SSIM` . + +| Method | Set5 PSNR | Set14 PSNR | DIV2K PSNR | Set5 SSIM | Set14 SSIM | DIV2K SSIM | GPU Info | Download | +| :----------------------------------------------------------------: | :-------: | :--------: | :--------: | :-------: | :--------: | :--------: | :------: | :------------------------------------------------------------------: | +| [swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py) | 38.3240 | 34.1174 | 37.8921 | 0.9626 | 0.9230 | 0.9481 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth) \| log | +| [swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py) | 34.8640 | 30.7669 | 34.1397 | 0.9317 | 0.8508 | 0.8917 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth) \| log | +| [swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py) | 32.7315 | 28.9065 | 32.0953 | 0.9029 | 0.7915 | 0.8418 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth) \| log | +| [swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k](/configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py) | 38.3971 | 34.4149 | 37.9473 | 0.9629 | 0.9252 | 0.9488 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth) \| log | +| [swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k](/configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py) | 34.9335 | 30.9258 | 34.2830 | 0.9323 | 0.8540 | 0.8939 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth) \| log | +| [swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k](/configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py) | 32.9214 | 29.0792 | 32.3021 | 0.9053 | 0.7953 | 0.8451 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth) \| log | + +### **Lightweight Image Super-Resolution** + +Evaluated on Y channels, `scale` pixels in each border are cropped before evaluation. +The metrics are `PSNR / SSIM` . + +| Method | Set5 PSNR | Set14 PSNR | DIV2K PSNR | Set5 SSIM | Set14 SSIM | DIV2K SSIM | GPU Info | Download | +| :----------------------------------------------------------------: | :-------: | :--------: | :--------: | :-------: | :--------: | :--------: | :------: | :------------------------------------------------------------------: | +| [swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py) | 38.1289 | 33.8404 | 37.5844 | 0.9617 | 0.9207 | 0.9459 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth) \| log | +| [swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py) | 34.6037 | 30.5340 | 33.8394 | 0.9293 | 0.8468 | 0.8867 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth) \| log | +| [swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py) | 32.4343 | 28.7441 | 31.8636 | 0.8984 | 0.7861 | 0.8353 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth) \| log | + +### **Real-World Image Super-Resolution** + +Evaluated on Y channels. +The metrics are `NIQE` . + +| Method | RealSRSet+5images NIQE | GPU Info | Download | +| :----------------------------------------------------------------------------------: | :--------------------: | :------: | :------------------------------------------------------------------------------------: | +| [swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 5.7975 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth) \| log | +| [swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 7.2738 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth) \| log | +| [swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 5.2329 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth) \| log | +| [swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 7.7460 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth) \| log | +| [swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py) | 5.1464 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/) \| log | +| [swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py) | 7.6378 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth) \| log | + +### **Grayscale Image Deoising** + +Evaluated on grayscale images. +The metrics are `PSNR` . + +| Method | Set12 PSNR | BSD68 PSNR | Urban100 PSNR | GPU Info | Download | +| :----------------------------------------------------------------------------: | :--------: | :--------: | :-----------: | :------: | :-------------------------------------------------------------------------------: | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py) | 33.9731 | 32.5203 | 34.3424 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py) | 31.6434 | 30.1377 | 31.9493 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py) | 28.5651 | 27.3157 | 28.6626 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth) \| log | + +### **Color Image Deoising** + +Evaluated on RGB channels. +The metrics are `PSNR` . + +| Method | CBSD68 PSNR | Kodak24 PSNR | McMaster PSNR | Urban100 PSNR | GPU Info | Download | +| :--------------------------------------------------------------------: | :---------: | :----------: | :-----------: | :-----------: | :------: | :-----------------------------------------------------------------------: | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py) | 34.4136 | 35.3555 | 35.6205 | 35.1836 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py) | 31.7626 | 32.9003 | 33.3198 | 32.9458 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py) | 28.5346 | 29.8058 | 30.2027 | 29.8832 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth) \| log | + +### **JPEG Compression Artifact Reduction (grayscale)** + +Evaluated on grayscale images. +The metrics are `PSNR / SSIM` . + +| Method | Classic5 PSNR | Classic5 SSIM | LIVE1 PSNR | LIVE1 SSIM | GPU Info | Download | +| :----------------------------------------------------------------------: | :-----------: | :-----------: | :--------: | :--------: | :------: | :------------------------------------------------------------------------: | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py) | 30.2746 | 0.8254 | 29.8611 | 0.8292 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py) | 32.5331 | 0.8753 | 32.2667 | 0.8914 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py) | 33.7504 | 0.8966 | 33.7001 | 0.9179 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py) | 34.5377 | 0.9087 | 34.6846 | 0.9322 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth) \| log | + +### **JPEG Compression Artifact Reduction (color)** + +Evaluated on RGB channels. +The metrics are `PSNR / SSIM` . + +| Method | Classic5 PSNR | Classic5 SSIM | LIVE1 PSNR | LIVE1 SSIM | GPU Info | Download | +| :----------------------------------------------------------------------: | :-----------: | :-----------: | :--------: | :--------: | :------: | :------------------------------------------------------------------------: | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py) | 30.1019 | 0.8217 | 28.0676 | 0.8094 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py) | 32.3489 | 0.8727 | 30.4514 | 0.8745 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py) | 33.6028 | 0.8949 | 31.8235 | 0.9023 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py) | 34.4344 | 0.9076 | 32.7610 | 0.9179 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40-5b77a6e6.pth) \| log | + +## Quick Start + +**Train** + +
+Train Instructions + +You can use the following commands to train a model with cpu or single/multiple GPUs. + +```shell +# cpu train +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py + +# 002 Lightweight Image Super-Resolution (small size) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py + +# 003 Real-World Image Super-Resolution +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py + +# 004 Grayscale Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py + +# 005 Color Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py + +# color +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py + + +# single-gpu train +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +python tools/train.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +python tools/train.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +python tools/train.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +python tools/train.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py + +# 002 Lightweight Image Super-Resolution (small size) +python tools/train.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py + +# 003 Real-World Image Super-Resolution +python tools/train.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py + +# 004 Grayscale Image Deoising (middle size) +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py + +# 005 Color Image Deoising (middle size) +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py + +# color +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py + + +# multi-gpu train +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +./tools/dist_train.sh configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py 8 + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +./tools/dist_train.sh configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py 8 + +# 002 Lightweight Image Super-Resolution (small size) +./tools/dist_train.sh configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py 8 + +# 003 Real-World Image Super-Resolution +./tools/dist_train.sh configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py 8 + +# 004 Grayscale Image Deoising (middle size) +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py 8 + +# 005 Color Image Deoising (middle size) +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py 8 + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py 8 + +# color +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py 8 +``` + +For more details, you can refer to **Train a model** part in [train_test.md](/docs/en/user_guides/train_test.md#Train-a-model-in-MMEditing). + +
+ +**Test** + +
+Test Instructions + +You can use the following commands to test a model with cpu or single/multiple GPUs. + +```shell +# cpu test +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth + + +# 002 Lightweight Image Super-Resolution (small size) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth + +# 003 Real-World Image Super-Resolution +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-25f1722a.pth + +# 004 Grayscale Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth + +# 005 Color Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding usesx8 blocks) +# grayscale +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth + + +# color +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40-5b77a6e6.pth + + + +# single-gpu test +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +python tools/test.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth + +python tools/test.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth + +python tools/test.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +python tools/test.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth + +python tools/test.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth + +python tools/test.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth + + +# 002 Lightweight Image Super-Resolution (small size) +python tools/test.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth + +python tools/test.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth + +python tools/test.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth + + +# 003 Real-World Image Super-Resolution +python tools/test.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth + +python tools/test.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth + +python tools/test.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth + +python tools/test.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth + +python tools/test.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth + +python tools/test.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-25f1722a.pth + + +# 004 Grayscale Image Deoising (middle size) +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth + + +# 005 Color Image Deoising (middle size) +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth + + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding usesx8 blocks) +# grayscale +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth + + +# color +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40-5b77a6e6.pth + + + +# multi-gpu test +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +./tools/dist_test.sh configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth + +./tools/dist_test.sh configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth + +./tools/dist_test.sh configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +./tools/dist_test.sh configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth + +./tools/dist_test.sh configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth + +./tools/dist_test.sh configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth + +# 002 Lightweight Image Super-Resolution (small size) +./tools/dist_test.sh configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth + +./tools/dist_test.sh configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth + +./tools/dist_test.sh configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth + +# 003 Real-World Image Super-Resolution +./tools/dist_test.sh configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth + +./tools/dist_test.sh configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth + +./tools/dist_test.sh configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth + +./tools/dist_test.sh configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth + +./tools/dist_test.sh configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth + +./tools/dist_test.sh configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-25f1722a.pth + +# 004 Grayscale Image Deoising (middle size) +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth + +# 005 Color Image Deoising (middle size) +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth + +# color +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py https://download.openmmlab.com/mmediting/swinir/ + +``` + +For more details, you can refer to **Test a pre-trained model** part in [train_test.md](/docs/en/user_guides/train_test.md#Test-a-pre-trained-model-in-MMEditing). + +
+ +## Citation + +```bibtex +@inproceedings{liang2021swinir, + title={Swinir: Image restoration using swin transformer}, + author={Liang, Jingyun and Cao, Jiezhang and Sun, Guolei and Zhang, Kai and Van Gool, Luc and Timofte, Radu}, + booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision}, + pages={1833--1844}, + year={2021} +} +``` diff --git a/configs/swinir/README_zh-CN.md b/configs/swinir/README_zh-CN.md new file mode 100644 index 0000000000..4fccfcbaac --- /dev/null +++ b/configs/swinir/README_zh-CN.md @@ -0,0 +1,503 @@ +# SwinIR (ICCVW'2021) + +> **任务**: 图像超分辨率, 图像去噪, JPEG压缩伪影移除 + + + +
+SwinIR (ICCVW'2021) + +```bibtex +@inproceedings{liang2021swinir, + title={Swinir: Image restoration using swin transformer}, + author={Liang, Jingyun and Cao, Jiezhang and Sun, Guolei and Zhang, Kai and Van Gool, Luc and Timofte, Radu}, + booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision}, + pages={1833--1844}, + year={2021} +} +``` + +
+ +
+ +### **Classical Image Super-Resolution** + +在 Y 通道上进行评估，在评估之前裁剪每个边界中的 `scale` 像素。 +我们使用 `PSNR` 和 `SSIM` 作为指标。 + +| 算法 | Set5 PSNR | Set14 PSNR | DIV2K PSNR | Set5 SSIM | Set14 SSIM | DIV2K SSIM | GPU 信息 | 下载 | +| :-----------------------------------------------------------------: | :-------: | :--------: | :--------: | :-------: | :--------: | :--------: | :------: | :-----------------------------------------------------------------: | +| [swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py) | 38.3240 | 34.1174 | 37.8921 | 0.9626 | 0.9230 | 0.9481 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth) \| log | +| [swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py) | 34.8640 | 30.7669 | 34.1397 | 0.9317 | 0.8508 | 0.8917 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth) \| log | +| [swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py) | 32.7315 | 28.9065 | 32.0953 | 0.9029 | 0.7915 | 0.8418 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth) \| log | +| [swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k](/configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py) | 38.3971 | 34.4149 | 37.9473 | 0.9629 | 0.9252 | 0.9488 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth) \| log | +| [swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k](/configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py) | 34.9335 | 30.9258 | 34.2830 | 0.9323 | 0.8540 | 0.8939 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth) \| log | +| [swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k](/configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py) | 32.9214 | 29.0792 | 32.3021 | 0.9053 | 0.7953 | 0.8451 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth) \| log | + +### **Lightweight Image Super-Resolution** + +在 Y 通道上进行评估，在评估之前裁剪每个边界中的 `scale` 像素。 +我们使用 `PSNR` 和 `SSIM` 作为指标。 + +| 算法 | Set5 PSNR | Set14 PSNR | DIV2K PSNR | Set5 SSIM | Set14 SSIM | DIV2K SSIM | GPU 信息 | 下载 | +| :-----------------------------------------------------------------: | :-------: | :--------: | :--------: | :-------: | :--------: | :--------: | :------: | :-----------------------------------------------------------------: | +| [swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py) | 38.1289 | 33.8404 | 37.5844 | 0.9617 | 0.9207 | 0.9459 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth) \| log | +| [swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py) | 34.6037 | 30.5340 | 33.8394 | 0.9293 | 0.8468 | 0.8867 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth) \| log | +| [swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k](/configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py) | 32.4343 | 28.7441 | 31.8636 | 0.8984 | 0.7861 | 0.8353 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth) \| log | + +### **Real-World Image Super-Resolution** + +在 Y 通道上进行评估。 +我们使用 NIQE 作为指标。 + +| 算法 | RealSRSet+5images NIQE | GPU 信息 | 下载 | +| :-----------------------------------------------------------------------------------: | :--------------------: | :------: | :-----------------------------------------------------------------------------------: | +| [swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 5.7975 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth) \| log | +| [swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 7.2738 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth) \| log | +| [swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 5.2329 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth) \| log | +| [swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py) | 7.7460 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth) \| log | +| [swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py) | 5.1464 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/) \| log | +| [swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost](/configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py) | 7.6378 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-25f1722a.pth) \| log | + +### **Grayscale Image Deoising** + +在灰度图上进行评估。 +我们使用 PSNR 作为指标。 + +| 算法 | Set12 PSNR | BSD68 PSNR | Urban100 PSNR | GPU 信息 | 下载 | +| :-----------------------------------------------------------------------------: | :--------: | :--------: | :-----------: | :------: | :------------------------------------------------------------------------------: | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py) | 33.9731 | 32.5203 | 34.3424 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py) | 31.6434 | 30.1377 | 31.9493 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py) | 28.5651 | 27.3157 | 28.6626 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth) \| log | + +### **Color Image Deoising** + +在 RGB 通道上进行评估。 +我们使用 PSNR 作为指标。 + +| 算法 | CBSD68 PSNR | Kodak24 PSNR | McMaster PSNR | Urban100 PSNR | GPU 信息 | 下载 | +| :---------------------------------------------------------------------: | :---------: | :----------: | :-----------: | :-----------: | :------: | :----------------------------------------------------------------------: | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py) | 34.4136 | 35.3555 | 35.6205 | 35.1836 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py) | 31.7626 | 32.9003 | 33.3198 | 32.9458 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth) \| log | +| [swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50](/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py) | 28.5346 | 29.8058 | 30.2027 | 29.8832 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth) \| log | + +### **JPEG Compression Artifact Reduction (grayscale)** + +在灰度图上进行评估。 +我们使用 PSNR 和 SSIM 作为指标。 + +| 算法 | Classic5 PSNR | Classic5 SSIM | LIVE1 PSNR | LIVE1 SSIM | GPU 信息 | 下载 | +| :-----------------------------------------------------------------------: | :-----------: | :-----------: | :--------: | :--------: | :------: | :-----------------------------------------------------------------------: | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py) | 30.2746 | 0.8254 | 29.8611 | 0.8292 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py) | 32.5331 | 0.8753 | 32.2667 | 0.8914 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py) | 33.7504 | 0.8966 | 33.7001 | 0.9179 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py) | 34.5377 | 0.9087 | 34.6846 | 0.9322 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth) \| log | + +### **JPEG Compression Artifact Reduction (color)** + +在 RGB 通道上进行评估。 +我们使用 PSNR 和 SSIM 作为指标。 + +| 算法 | Classic5 PSNR | Classic5 SSIM | LIVE1 PSNR | LIVE1 SSIM | GPU 信息 | 下载 | +| :-----------------------------------------------------------------------: | :-----------: | :-----------: | :--------: | :--------: | :------: | :-----------------------------------------------------------------------: | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py) | 30.1019 | 0.8217 | 28.0676 | 0.8094 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py) | 32.3489 | 0.8727 | 30.3489 | 0.8745 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py) | 33.6028 | 0.8949 | 31.8235 | 0.9023 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth) \| log | +| [swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40](/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py) | 34.4344 | 0.9076 | 32.7610 | 0.9179 | 8 | [model](https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40-5b77a6e6.pth) \| log | + +## 快速开始 + +**训练** + +
+训练说明 + +您可以使用以下命令来训练模型。 + +```shell +# CPU上训练 +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py + +# 002 Lightweight Image Super-Resolution (small size) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py + +# 003 Real-World Image Super-Resolution +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py + +# 004 Grayscale Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py + +# 005 Color Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py + +# color +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py +CUDA_VISIBLE_DEVICES=-1 python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py + + + +# 单个GPU上训练 +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +python tools/train.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +python tools/train.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +python tools/train.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py +python tools/train.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py + +# 002 Lightweight Image Super-Resolution (small size) +python tools/train.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py +python tools/train.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py + +# 003 Real-World Image Super-Resolution +python tools/train.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py +python tools/train.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py + +# 004 Grayscale Image Deoising (middle size) +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py + +# 005 Color Image Deoising (middle size) +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py +python tools/train.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py + +# color +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py +python tools/train.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py + + + +# 多个GPU上训练 +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +./tools/dist_train.sh configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py 8 + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +./tools/dist_train.sh configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py 8 + +# 002 Lightweight Image Super-Resolution (small size) +./tools/dist_train.sh configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py 8 +./tools/dist_train.sh configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py 8 + +# 003 Real-World Image Super-Resolution +./tools/dist_train.sh configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py 8 +./tools/dist_train.sh configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py 8 + +# 004 Grayscale Image Deoising (middle size) +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py 8 + +# 005 Color Image Deoising (middle size) +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py 8 +./tools/dist_train.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py 8 + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py 8 + +# color +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py 8 +./tools/dist_train.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py 8 +``` + +更多细节可以参考 [train_test.md](/docs/zh_cn/user_guides/train_test.md) 中的 **Train a model** 部分。 + +
+ +**测试** + +
+测试说明 + +您可以使用以下命令来测试模型。 + +```shell +# CPU上测试 +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth + + +# 002 Lightweight Image Super-Resolution (small size) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth + +# 003 Real-World Image Super-Resolution +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-25f1722a.pth + +# 004 Grayscale Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth + +# 005 Color Image Deoising (middle size) +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding usesx8 blocks) +# grayscale +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth + + +# color +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth + +CUDA_VISIBLE_DEVICES=-1 python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40-5b77a6e6.pth + + + +# 单个GPU上测试 +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +python tools/test.py configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth + +python tools/test.py configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth + +python tools/test.py configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +python tools/test.py configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth + +python tools/test.py configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth + +python tools/test.py configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth + + +# 002 Lightweight Image Super-Resolution (small size) +python tools/test.py configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth + +python tools/test.py configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth + +python tools/test.py configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth + + +# 003 Real-World Image Super-Resolution +python tools/test.py configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth + +python tools/test.py configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth + +python tools/test.py configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth + +python tools/test.py configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth + +python tools/test.py configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth + +python tools/test.py configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-25f1722a.pth + + +# 004 Grayscale Image Deoising (middle size) +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth + + +# 005 Color Image Deoising (middle size) +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth + +python tools/test.py configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth + + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding usesx8 blocks) +# grayscale +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth + + +# color +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth + +python tools/test.py configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40-5b77a6e6.pth + + + +# 多GPU测试 +# 001 Classical Image Super-Resolution (middle size) +# (setting1: when model is trained on DIV2K and with training_patch_size=48) +./tools/dist_test.sh configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth + +./tools/dist_test.sh configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth + +./tools/dist_test.sh configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth + +# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64) +./tools/dist_test.sh configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth + +./tools/dist_test.sh configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth + +./tools/dist_test.sh configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth + +# 002 Lightweight Image Super-Resolution (small size) +./tools/dist_test.sh configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth + +./tools/dist_test.sh configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth + +./tools/dist_test.sh configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth + +# 003 Real-World Image Super-Resolution +./tools/dist_test.sh configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth + +./tools/dist_test.sh configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth + +./tools/dist_test.sh configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth + +./tools/dist_test.sh configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth + +./tools/dist_test.sh configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth + +./tools/dist_test.sh configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-25f1722a.pth + +# 004 Grayscale Image Deoising (middle size) +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth + +# 005 Color Image Deoising (middle size) +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth + +./tools/dist_test.sh configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth + +# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks) +# grayscale +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth + +# color +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth + +./tools/dist_test.sh configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py https://download.openmmlab.com/mmediting/swinir/ + +``` + +更多细节可以参考 [train_test.md](/docs/zh_cn/user_guides/train_test.md) 中的 **Test a pre-trained model** 部分。 + +
diff --git a/configs/swinir/metafile.yml b/configs/swinir/metafile.yml new file mode 100644 index 0000000000..f6a4a58617 --- /dev/null +++ b/configs/swinir/metafile.yml @@ -0,0 +1,446 @@ +Collections: +- Metadata: + Architecture: + - SwinIR + Name: SwinIR + Paper: + - https://arxiv.org/abs/2108.10257 + README: configs/swinir/README.md + Task: + - image super-resolution + - image denoising + - jpeg compression artifact reduction + Year: 2021 +Models: +- Config: configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: DIV2K + Name: swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k + Results: + - Dataset: DIV2K + Metrics: + DIV2K PSNR: 37.8921 + DIV2K SSIM: 0.9481 + Set14 PSNR: 34.1174 + Set14 SSIM: 0.923 + Set5 PSNR: 38.324 + Set5 SSIM: 0.9626 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k-ed2d419e.pth +- Config: configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: DIV2K + Name: swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k + Results: + - Dataset: DIV2K + Metrics: + DIV2K PSNR: 34.1397 + DIV2K SSIM: 0.8917 + Set14 PSNR: 30.7669 + Set14 SSIM: 0.8508 + Set5 PSNR: 34.864 + Set5 SSIM: 0.9317 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k-926950f1.pth +- Config: configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: DIV2K + Name: swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k + Results: + - Dataset: DIV2K + Metrics: + DIV2K PSNR: 32.0953 + DIV2K SSIM: 0.8418 + Set14 PSNR: 28.9065 + Set14 SSIM: 0.7915 + Set5 PSNR: 32.7315 + Set5 SSIM: 0.9029 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k-88e4903d.pth +- Config: configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k + Results: + - Dataset: Others + Metrics: + DIV2K PSNR: 37.9473 + DIV2K SSIM: 0.9488 + Set14 PSNR: 34.4149 + Set14 SSIM: 0.9252 + Set5 PSNR: 38.3971 + Set5 SSIM: 0.9629 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k-69e15fb6.pth +- Config: configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k + Results: + - Dataset: Others + Metrics: + DIV2K PSNR: 34.283 + DIV2K SSIM: 0.8939 + Set14 PSNR: 30.9258 + Set14 SSIM: 0.854 + Set5 PSNR: 34.9335 + Set5 SSIM: 0.9323 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k-d6982f7b.pth +- Config: configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k + Results: + - Dataset: Others + Metrics: + DIV2K PSNR: 32.3021 + DIV2K SSIM: 0.8451 + Set14 PSNR: 29.0792 + Set14 SSIM: 0.7953 + Set5 PSNR: 32.9214 + Set5 SSIM: 0.9053 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k-0502d775.pth +- Config: configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: DIV2K + Name: swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k + Results: + - Dataset: DIV2K + Metrics: + DIV2K PSNR: 37.5844 + DIV2K SSIM: 0.9459 + Set14 PSNR: 33.8404 + Set14 SSIM: 0.9207 + Set5 PSNR: 38.1289 + Set5 SSIM: 0.9617 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k-131d3f64.pth +- Config: configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: DIV2K + Name: swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k + Results: + - Dataset: DIV2K + Metrics: + DIV2K PSNR: 33.8394 + DIV2K SSIM: 0.8867 + Set14 PSNR: 30.534 + Set14 SSIM: 0.8468 + Set5 PSNR: 34.6037 + Set5 SSIM: 0.9293 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k-309cb239.pth +- Config: configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: DIV2K + Name: swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k + Results: + - Dataset: DIV2K + Metrics: + DIV2K PSNR: 31.8636 + DIV2K SSIM: 0.8353 + Set14 PSNR: 28.7441 + Set14 SSIM: 0.7861 + Set5 PSNR: 32.4343 + Set5 SSIM: 0.8984 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k-d6622d03.pth +- Config: configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost + Results: + - Dataset: Others + Metrics: + RealSRSet+5images NIQE: 5.7975 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-c6425057.pth +- Config: configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost + Results: + - Dataset: Others + Metrics: + RealSRSet+5images NIQE: 7.2738 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-6f0c425f.pth +- Config: configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost + Results: + - Dataset: Others + Metrics: + RealSRSet+5images NIQE: 5.2329 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-36960d18.pth +- Config: configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost + Results: + - Dataset: Others + Metrics: + RealSRSet+5images NIQE: 7.746 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-os-a016a72f.pth +- Config: configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost + Results: + - Dataset: Others + Metrics: + RealSRSet+5images NIQE: 5.1464 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/ +- Config: configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost + Results: + - Dataset: Others + Metrics: + RealSRSet+5images NIQE: 7.6378 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-os-9f1599b5.pth +- Config: configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15 + Results: + - Dataset: Others + Metrics: + BSD68 PSNR: 32.5203 + Set12 PSNR: 33.9731 + Urban100 PSNR: 34.3424 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15-6782691b.pth +- Config: configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25 + Results: + - Dataset: Others + Metrics: + BSD68 PSNR: 30.1377 + Set12 PSNR: 31.6434 + Urban100 PSNR: 31.9493 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25-d0d8d4da.pth +- Config: configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50 + Results: + - Dataset: Others + Metrics: + BSD68 PSNR: 27.3157 + Set12 PSNR: 28.5651 + Urban100 PSNR: 28.6626 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50-54c9968a.pth +- Config: configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15 + Results: + - Dataset: Others + Metrics: + CBSD68 PSNR: 34.4136 + Kodak24 PSNR: 35.3555 + McMaster PSNR: 35.6205 + Urban100 PSNR: 35.1836 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15-c74a2cee.pth +- Config: configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25 + Results: + - Dataset: Others + Metrics: + CBSD68 PSNR: 31.7626 + Kodak24 PSNR: 32.9003 + McMaster PSNR: 33.3198 + Urban100 PSNR: 32.9458 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25-df2b1c0c.pth +- Config: configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50 + Results: + - Dataset: Others + Metrics: + CBSD68 PSNR: 28.5346 + Kodak24 PSNR: 29.8058 + McMaster PSNR: 30.2027 + Urban100 PSNR: 29.8832 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50-e369874c.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 30.2746 + Classic5 SSIM: 0.8254 + LIVE1 PSNR: 29.8611 + LIVE1 SSIM: 0.8292 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10-da93c8e9.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 32.5331 + Classic5 SSIM: 0.8753 + LIVE1 PSNR: 32.2667 + LIVE1 SSIM: 0.8914 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20-d47367b1.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 33.7504 + Classic5 SSIM: 0.8966 + LIVE1 PSNR: 33.7001 + LIVE1 SSIM: 0.9179 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30-52c083cf.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 34.5377 + Classic5 SSIM: 0.9087 + LIVE1 PSNR: 34.6846 + LIVE1 SSIM: 0.9322 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40-803e8d9b.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 30.1019 + Classic5 SSIM: 0.8217 + LIVE1 PSNR: 28.0676 + LIVE1 SSIM: 0.8094 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10-09aafadc.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 32.3489 + Classic5 SSIM: 0.8727 + LIVE1 PSNR: 30.4514 + LIVE1 SSIM: 0.8745 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20-b8a42b5e.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 33.6028 + Classic5 SSIM: 0.8949 + LIVE1 PSNR: 31.8235 + LIVE1 SSIM: 0.9023 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30-e9fe6859.pth +- Config: configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py + In Collection: SwinIR + Metadata: + GPUs: '8' + Training Data: Others + Name: swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40 + Results: + - Dataset: Others + Metrics: + Classic5 PSNR: 34.4344 + Classic5 SSIM: 0.9076 + LIVE1 PSNR: 32.761 + LIVE1 SSIM: 0.9179 + Task: Image Super-Resolution, Image denoising, JPEG compression artifact reduction + Weights: https://download.openmmlab.com/mmediting/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40-5b77a6e6.pth diff --git a/configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py b/configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py new file mode 100644 index 0000000000..b527c7b70a --- /dev/null +++ b/configs/swinir/swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py @@ -0,0 +1,5 @@ +_base_ = ['swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py'] + +experiment_name = 'swinir_gan-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' diff --git a/configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py b/configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py new file mode 100644 index 0000000000..fff5e6b4a5 --- /dev/null +++ b/configs/swinir/swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py @@ -0,0 +1,5 @@ +_base_ = ['swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py'] + +experiment_name = 'swinir_gan-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' diff --git a/configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py b/configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py new file mode 100644 index 0000000000..86bd2a9175 --- /dev/null +++ b/configs/swinir/swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py @@ -0,0 +1,5 @@ +_base_ = ['swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py'] + +experiment_name = 'swinir_gan-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' diff --git a/configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py b/configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py new file mode 100644 index 0000000000..ee1dc77bfb --- /dev/null +++ b/configs/swinir/swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py @@ -0,0 +1,61 @@ +_base_ = ['../_base_/default_runtime.py'] + +experiment_name = 'swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 2 +img_size = 64 + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=scale, + in_chans=3, + img_size=img_size, + window_size=8, + img_range=1.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='nearest+conv', + resi_connection='1conv'), + pixel_loss=dict(type='L1Loss', loss_weight=1.0, reduction='mean'), + data_preprocessor=dict( + type='EditDataPreprocessor', mean=[0., 0., 0.], std=[255., 255., + 255.])) + +test_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='PackEditInputs') +] + +test_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type='BasicImageDataset', + metainfo=dict(dataset_type='realsrset', task_name='realsr'), + data_root='data/RealSRSet+5images', + data_prefix=dict(img='', gt=''), + pipeline=test_pipeline)) + +test_evaluator = [dict(type='NIQE', input_order='CHW', convert_to='Y')] + +test_cfg = dict(type='TestLoop') diff --git a/configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py b/configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py new file mode 100644 index 0000000000..2f8e0451ec --- /dev/null +++ b/configs/swinir/swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py @@ -0,0 +1,10 @@ +_base_ = ['swinir_psnr-x2s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py'] + +experiment_name = 'swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 4 + +# model settings +model = dict(generator=dict(upscale=scale)) diff --git a/configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py b/configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py new file mode 100644 index 0000000000..d84a851aa3 --- /dev/null +++ b/configs/swinir/swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost.py @@ -0,0 +1,13 @@ +_base_ = ['swinir_psnr-x4s64w8d6e180_8xb4-lr1e-4-600k_df2k-ost.py'] + +experiment_name = 'swinir_psnr-x4s64w8d9e240_8xb4-lr1e-4-600k_df2k-ost' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# model settings +model = dict( + generator=dict( + depths=[6, 6, 6, 6, 6, 6, 6, 6, 6], + embed_dim=240, + num_heads=[8, 8, 8, 8, 8, 8, 8, 8, 8], + resi_connection='3conv')) diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py new file mode 100644 index 0000000000..7eea2d978c --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py @@ -0,0 +1,134 @@ +_base_ = [ + '../_base_/default_runtime.py', + '../_base_/datasets/decompression_test_config.py' +] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +quality = 10 + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=1, + in_chans=3, + img_size=126, + window_size=7, + img_range=255.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='', + resi_connection='1conv'), + pixel_loss=dict(type='CharbonnierLoss', eps=1e-9), + data_preprocessor=dict( + type='EditDataPreprocessor', mean=[0., 0., 0.], std=[255., 255., + 255.])) + +train_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='SetValues', dictionary=dict(scale=1)), + dict(type='PairedRandomCrop', gt_patch_size=126), + dict( + type='Flip', + keys=['img', 'gt'], + flip_ratio=0.5, + direction='horizontal'), + dict( + type='Flip', keys=['img', 'gt'], flip_ratio=0.5, direction='vertical'), + dict(type='RandomTransposeHW', keys=['img', 'gt'], transpose_ratio=0.5), + dict( + type='RandomJPEGCompression', + params=dict(quality=[quality, quality], color_type='color'), + keys=['img']), + dict(type='PackEditInputs') +] + +val_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='RandomJPEGCompression', + params=dict(quality=[quality, quality], color_type='color'), + keys=['img']), + dict(type='PackEditInputs') +] + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=2, + batch_size=1, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DFWB8550sub_GT.txt', + metainfo=dict(dataset_type='dfwb', task_name='CAR'), + data_root=data_root + '/DFWB', + data_prefix=dict(img='', gt=''), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) + +val_dataloader = dict( + num_workers=2, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type=dataset_type, + metainfo=dict(dataset_type='live1', task_name='CAR'), + data_root=data_root + '/LIVE1', + data_prefix=dict(img='', gt=''), + pipeline=val_pipeline)) + +val_evaluator = [ + dict(type='PSNR', prefix='LIVE1'), + dict(type='SSIM', prefix='LIVE1'), +] + +train_cfg = dict( + type='IterBasedTrainLoop', max_iters=1_600_000, val_interval=5000) +val_cfg = dict(type='ValLoop') + +# optimizer +optim_wrapper = dict( + constructor='DefaultOptimWrapperConstructor', + type='OptimWrapper', + optimizer=dict(type='Adam', lr=2e-4, betas=(0.9, 0.999))) + +# learning policy +param_scheduler = dict( + type='MultiStepLR', + by_epoch=False, + milestones=[800000, 1200000, 1400000, 1500000, 1600000], + gamma=0.5) diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py new file mode 100644 index 0000000000..f83071f0d0 --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py'] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR20' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify JPEG quality factor of RandomJPEGCompression +quality = 20 +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['quality'] = [quality, quality] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['quality'] = [quality, quality] + +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['quality'] = [quality, quality] diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py new file mode 100644 index 0000000000..44872b9de8 --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py'] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR30' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify JPEG quality factor of RandomJPEGCompression +quality = 30 +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['quality'] = [quality, quality] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['quality'] = [quality, quality] + +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['quality'] = [quality, quality] diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py new file mode 100644 index 0000000000..69ae48385a --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR10.py'] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-colorCAR40' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify JPEG quality factor of RandomJPEGCompression +quality = 40 +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['quality'] = [quality, quality] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['quality'] = [quality, quality] + +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['quality'] = [quality, quality] diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py new file mode 100644 index 0000000000..d4206be1c5 --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py @@ -0,0 +1,139 @@ +_base_ = [ + '../_base_/default_runtime.py', + '../_base_/datasets/decompression_test_config.py' +] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +quality = 10 + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=1, + in_chans=1, + img_size=126, + window_size=7, + img_range=255.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='', + resi_connection='1conv'), + pixel_loss=dict(type='CharbonnierLoss', eps=1e-9), + data_preprocessor=dict(type='EditDataPreprocessor', mean=[0.], std=[255.])) + +train_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='grayscale', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='grayscale', + imdecode_backend='cv2'), + dict(type='SetValues', dictionary=dict(scale=1)), + dict(type='PairedRandomCrop', gt_patch_size=126), + dict( + type='Flip', + keys=['img', 'gt'], + flip_ratio=0.5, + direction='horizontal'), + dict( + type='Flip', keys=['img', 'gt'], flip_ratio=0.5, direction='vertical'), + dict(type='RandomTransposeHW', keys=['img', 'gt'], transpose_ratio=0.5), + dict( + type='RandomJPEGCompression', + params=dict(quality=[quality, quality], color_type='grayscale'), + keys=['img']), + dict(type='PackEditInputs') +] + +val_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='grayscale', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='grayscale', + imdecode_backend='cv2'), + dict( + type='RandomJPEGCompression', + params=dict(quality=[quality, quality], color_type='grayscale'), + keys=['img']), + dict(type='PackEditInputs') +] + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=1, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DFWB8550sub_GT.txt', + metainfo=dict(dataset_type='dfwb', task_name='CAR'), + data_root=data_root + '/DFWB', + data_prefix=dict(img='', gt=''), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) + +val_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type=dataset_type, + metainfo=dict(dataset_type='classic5', task_name='CAR'), + data_root=data_root + '/classic5', + data_prefix=dict(img='', gt=''), + pipeline=val_pipeline)) + +val_evaluator = [ + dict(type='PSNR', prefix='classic5'), + dict(type='SSIM', prefix='classic5'), +] + +train_cfg = dict( + type='IterBasedTrainLoop', max_iters=1_600_000, val_interval=5000) +val_cfg = dict(type='ValLoop') + +test_dataloader = _base_.test_dataloader +for idx in range(len(test_dataloader)): + test_pipeline = test_dataloader[idx]['dataset']['pipeline'] + if idx > 0: + test_pipeline[0]['to_y_channel'] = True + test_pipeline[1]['to_y_channel'] = True + else: + test_pipeline[0]['color_type'] = 'grayscale' + test_pipeline[1]['color_type'] = 'grayscale' + test_pipeline[2]['params']['color_type'] = 'grayscale' + +# optimizer +optim_wrapper = dict( + constructor='DefaultOptimWrapperConstructor', + type='OptimWrapper', + optimizer=dict(type='Adam', lr=2e-4, betas=(0.9, 0.999))) + +# learning policy +param_scheduler = dict( + type='MultiStepLR', + by_epoch=False, + milestones=[800000, 1200000, 1400000, 1500000, 1600000], + gamma=0.5) diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py new file mode 100644 index 0000000000..f3ea47b95f --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py'] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR20' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify JPEG quality factor of RandomJPEGCompression +quality = 20 +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['quality'] = [quality, quality] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['quality'] = [quality, quality] + +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['quality'] = [quality, quality] diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py new file mode 100644 index 0000000000..6054939a18 --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py'] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR30' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify JPEG quality factor of RandomJPEGCompression +quality = 30 +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['quality'] = [quality, quality] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['quality'] = [quality, quality] + +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['quality'] = [quality, quality] diff --git a/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py new file mode 100644 index 0000000000..4b0d6d6f02 --- /dev/null +++ b/configs/swinir/swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR10.py'] + +experiment_name = 'swinir_s126w7d6e180_8xb1-lr2e-4-1600k_dfwb-grayCAR40' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify JPEG quality factor of RandomJPEGCompression +quality = 40 +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['quality'] = [quality, quality] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['quality'] = [quality, quality] + +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['quality'] = [quality, quality] diff --git a/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py new file mode 100644 index 0000000000..469e4c6e8b --- /dev/null +++ b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py @@ -0,0 +1,147 @@ +_base_ = [ + '../_base_/default_runtime.py', + '../_base_/datasets/denoising-gaussian_color_test_config.py' +] + +experiment_name = 'swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify sigma of RandomNoise +sigma = 15 +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=1, + in_chans=3, + img_size=128, + window_size=8, + img_range=1.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='', + resi_connection='1conv'), + pixel_loss=dict(type='CharbonnierLoss', eps=1e-9), + data_preprocessor=dict( + type='EditDataPreprocessor', mean=[0., 0., 0.], std=[255., 255., + 255.])) + +train_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='SetValues', dictionary=dict(scale=1)), + dict(type='PairedRandomCrop', gt_patch_size=128), + dict( + type='Flip', + keys=['img', 'gt'], + flip_ratio=0.5, + direction='horizontal'), + dict( + type='Flip', keys=['img', 'gt'], flip_ratio=0.5, direction='vertical'), + dict(type='RandomTransposeHW', keys=['img', 'gt'], transpose_ratio=0.5), + dict( + type='RandomNoise', + params=dict( + noise_type=['gaussian'], + noise_prob=[1], + gaussian_sigma=[sigma * 255, sigma * 255], + gaussian_gray_noise_prob=0), + keys=['img']), + dict(type='PackEditInputs') +] + +val_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='RandomNoise', + params=dict( + noise_type=['gaussian'], + noise_prob=[1], + gaussian_sigma=[sigma * 255, sigma * 255], + gaussian_gray_noise_prob=0), + keys=['img']), + dict(type='PackEditInputs') +] + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=1, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DFWB8550sub_GT.txt', + metainfo=dict(dataset_type='dfwb', task_name='denoising'), + data_root=data_root + '/DFWB', + data_prefix=dict(img='', gt=''), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) + +val_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type=dataset_type, + metainfo=dict(dataset_type='mcmaster', task_name='denoising'), + data_root=data_root + '/McMaster', + data_prefix=dict(img='', gt=''), + pipeline=val_pipeline)) + +val_evaluator = [ + dict(type='PSNR', prefix='McMaster'), + dict(type='SSIM', prefix='McMaster'), +] + +train_cfg = dict( + type='IterBasedTrainLoop', max_iters=1_600_000, val_interval=5000) +val_cfg = dict(type='ValLoop') + +# optimizer +optim_wrapper = dict( + constructor='DefaultOptimWrapperConstructor', + type='OptimWrapper', + optimizer=dict(type='Adam', lr=2e-4, betas=(0.9, 0.999))) + +# learning policy +param_scheduler = dict( + type='MultiStepLR', + by_epoch=False, + milestones=[800000, 1200000, 1400000, 1500000, 1600000], + gamma=0.5) diff --git a/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py new file mode 100644 index 0000000000..1b11cb3c80 --- /dev/null +++ b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py'] + +experiment_name = 'swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN25' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify sigma of RandomNoise +sigma = 25 +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] + +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['gaussian_sigma'] = [sigma, sigma] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] diff --git a/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py new file mode 100644 index 0000000000..c57d636a71 --- /dev/null +++ b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN15.py'] + +experiment_name = 'swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-colorDN50' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify sigma of RandomNoise +sigma = 50 +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] + +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['gaussian_sigma'] = [sigma, sigma] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] diff --git a/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py new file mode 100644 index 0000000000..2f765c3c3a --- /dev/null +++ b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py @@ -0,0 +1,141 @@ +_base_ = [ + '../_base_/default_runtime.py', + '../_base_/datasets/denoising-gaussian_gray_test_config.py' +] + +experiment_name = 'swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify sigma of RandomNoise +sigma = 15 +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=1, + in_chans=1, + img_size=128, + window_size=8, + img_range=1.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='', + resi_connection='1conv'), + pixel_loss=dict(type='CharbonnierLoss', eps=1e-9), + data_preprocessor=dict(type='EditDataPreprocessor', mean=[0.], std=[255.])) + +train_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='grayscale', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='grayscale', + imdecode_backend='cv2'), + dict(type='SetValues', dictionary=dict(scale=1)), + dict(type='PairedRandomCrop', gt_patch_size=128), + dict( + type='Flip', + keys=['img', 'gt'], + flip_ratio=0.5, + direction='horizontal'), + dict( + type='Flip', keys=['img', 'gt'], flip_ratio=0.5, direction='vertical'), + dict(type='RandomTransposeHW', keys=['img', 'gt'], transpose_ratio=0.5), + dict( + type='RandomNoise', + params=dict( + noise_type=['gaussian'], + noise_prob=[1], + gaussian_sigma=[sigma, sigma], + gaussian_gray_noise_prob=0), + keys=['img']), + dict(type='PackEditInputs') +] + +val_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='grayscale', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='grayscale', + imdecode_backend='cv2'), + dict( + type='RandomNoise', + params=dict( + noise_type=['gaussian'], + noise_prob=[1], + gaussian_sigma=[sigma, sigma], + gaussian_gray_noise_prob=0), + keys=['img']), + dict(type='PackEditInputs') +] + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=1, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DFWB8550sub_GT.txt', + metainfo=dict(dataset_type='dfwb', task_name='denoising'), + data_root=data_root + '/DFWB', + data_prefix=dict(img='', gt=''), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) + +val_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type=dataset_type, + metainfo=dict(dataset_type='set12', task_name='denoising'), + data_root=data_root + '/Set12', + data_prefix=dict(img='', gt=''), + pipeline=val_pipeline)) + +val_evaluator = [ + dict(type='PSNR', prefix='Set12'), + dict(type='SSIM', prefix='Set12'), +] + +train_cfg = dict( + type='IterBasedTrainLoop', max_iters=1_600_000, val_interval=5000) +val_cfg = dict(type='ValLoop') + +# optimizer +optim_wrapper = dict( + constructor='DefaultOptimWrapperConstructor', + type='OptimWrapper', + optimizer=dict(type='Adam', lr=2e-4, betas=(0.9, 0.999))) + +# learning policy +param_scheduler = dict( + type='MultiStepLR', + by_epoch=False, + milestones=[800000, 1200000, 1400000, 1500000, 1600000], + gamma=0.5) diff --git a/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py new file mode 100644 index 0000000000..34f364400b --- /dev/null +++ b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py'] + +experiment_name = 'swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN25' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify sigma of RandomNoise +sigma = 25 +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] + +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['gaussian_sigma'] = [sigma, sigma] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] diff --git a/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py new file mode 100644 index 0000000000..5f96a5ae6d --- /dev/null +++ b/configs/swinir/swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50.py @@ -0,0 +1,20 @@ +_base_ = ['swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN15.py'] + +experiment_name = 'swinir_s128w8d6e180_8xb1-lr2e-4-1600k_dfwb-grayDN50' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +# modify sigma of RandomNoise +sigma = 50 +test_dataloader = _base_.test_dataloader +for dataloader in test_dataloader: + test_pipeline = dataloader['dataset']['pipeline'] + test_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] + +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[-2]['params']['gaussian_sigma'] = [sigma, sigma] + +val_dataloader = _base_.val_dataloader +val_pipeline = val_dataloader['dataset']['pipeline'] +val_pipeline[2]['params']['gaussian_sigma'] = [sigma, sigma] diff --git a/configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py b/configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py new file mode 100644 index 0000000000..7abface52f --- /dev/null +++ b/configs/swinir/swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py @@ -0,0 +1,133 @@ +_base_ = [ + '../_base_/default_runtime.py', '../_base_/datasets/sisr_x2_test_config.py' +] + +experiment_name = 'swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 2 +img_size = 48 + +# evaluated on Y channels +test_evaluator = _base_.test_evaluator +for evaluator in test_evaluator: + for metric in evaluator: + metric['convert_to'] = 'Y' + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=scale, + in_chans=3, + img_size=img_size, + window_size=8, + img_range=1.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='pixelshuffle', + resi_connection='1conv'), + pixel_loss=dict(type='L1Loss', loss_weight=1.0, reduction='mean'), + data_preprocessor=dict( + type='EditDataPreprocessor', mean=[0., 0., 0.], std=[255., 255., + 255.])) + +train_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='SetValues', dictionary=dict(scale=scale)), + dict(type='PairedRandomCrop', gt_patch_size=img_size * scale), + dict( + type='Flip', + keys=['img', 'gt'], + flip_ratio=0.5, + direction='horizontal'), + dict( + type='Flip', keys=['img', 'gt'], flip_ratio=0.5, direction='vertical'), + dict(type='RandomTransposeHW', keys=['img', 'gt'], transpose_ratio=0.5), + dict(type='PackEditInputs') +] + +val_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='PackEditInputs') +] + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=4, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DIV2K800sub_GT.txt', + metainfo=dict(dataset_type='div2k', task_name='sisr'), + data_root=data_root + '/DIV2K', + data_prefix=dict( + img='DIV2K_train_LR_bicubic/X2_sub', gt='DIV2K_train_HR_sub'), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) + +val_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type=dataset_type, + metainfo=dict(dataset_type='set5', task_name='sisr'), + data_root=data_root + '/Set5', + data_prefix=dict(img='LRbicx2', gt='GTmod12'), + pipeline=val_pipeline)) + +val_evaluator = [ + dict(type='PSNR', crop_border=scale), + dict(type='SSIM', crop_border=scale), +] + +train_cfg = dict( + type='IterBasedTrainLoop', max_iters=500_000, val_interval=5000) +val_cfg = dict(type='ValLoop') + +# optimizer +optim_wrapper = dict( + constructor='DefaultOptimWrapperConstructor', + type='OptimWrapper', + optimizer=dict(type='Adam', lr=2e-4, betas=(0.9, 0.999))) + +# learning policy +param_scheduler = dict( + type='MultiStepLR', + by_epoch=False, + milestones=[250000, 400000, 450000, 475000], + gamma=0.5) diff --git a/configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py b/configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py new file mode 100644 index 0000000000..88f5d001bf --- /dev/null +++ b/configs/swinir/swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k.py @@ -0,0 +1,22 @@ +_base_ = ['swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py'] + +experiment_name = 'swinir_x2s64w8d4e60_8xb4-lr2e-4-500k_div2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 2 +img_size = 64 + +# model settings +model = dict( + generator=dict( + img_size=img_size, + depths=[6, 6, 6, 6], + embed_dim=60, + num_heads=[6, 6, 6, 6], + upsampler='pixelshuffledirect')) + +# modify patch size of train_dataloader +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[3]['gt_patch_size'] = img_size * scale diff --git a/configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py b/configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py new file mode 100644 index 0000000000..475e42eef5 --- /dev/null +++ b/configs/swinir/swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k.py @@ -0,0 +1,35 @@ +_base_ = ['swinir_x2s48w8d6e180_8xb4-lr2e-4-500k_div2k.py'] + +experiment_name = 'swinir_x2s64w8d6e180_8xb4-lr2e-4-500k_df2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 2 +img_size = 64 + +# model settings +model = dict(generator=dict(img_size=img_size)) + +# modify patch size of train_pipeline +train_pipeline = _base_.train_pipeline +train_pipeline[3]['gt_patch_size'] = img_size * scale + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=4, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DF2K3450sub_GT.txt', + metainfo=dict(dataset_type='div2k', task_name='sisr'), + data_root=data_root + '/DF2K', + data_prefix=dict( + img='DF2K_train_LR_bicubic/X2_sub', gt='DF2K_train_HR_sub'), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) diff --git a/configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py b/configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py new file mode 100644 index 0000000000..20f8e1c533 --- /dev/null +++ b/configs/swinir/swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py @@ -0,0 +1,133 @@ +_base_ = [ + '../_base_/default_runtime.py', '../_base_/datasets/sisr_x3_test_config.py' +] + +experiment_name = 'swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 3 +img_size = 48 + +# evaluated on Y channels +test_evaluator = _base_.test_evaluator +for evaluator in test_evaluator: + for metric in evaluator: + metric['convert_to'] = 'Y' + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=scale, + in_chans=3, + img_size=img_size, + window_size=8, + img_range=1.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='pixelshuffle', + resi_connection='1conv'), + pixel_loss=dict(type='L1Loss', loss_weight=1.0, reduction='mean'), + data_preprocessor=dict( + type='EditDataPreprocessor', mean=[0., 0., 0.], std=[255., 255., + 255.])) + +train_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='SetValues', dictionary=dict(scale=scale)), + dict(type='PairedRandomCrop', gt_patch_size=img_size * scale), + dict( + type='Flip', + keys=['img', 'gt'], + flip_ratio=0.5, + direction='horizontal'), + dict( + type='Flip', keys=['img', 'gt'], flip_ratio=0.5, direction='vertical'), + dict(type='RandomTransposeHW', keys=['img', 'gt'], transpose_ratio=0.5), + dict(type='PackEditInputs') +] + +val_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='PackEditInputs') +] + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=4, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DIV2K800sub_GT.txt', + metainfo=dict(dataset_type='div2k', task_name='sisr'), + data_root=data_root + '/DIV2K', + data_prefix=dict( + img='DIV2K_train_LR_bicubic/X3_sub', gt='DIV2K_train_HR_sub'), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) + +val_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type=dataset_type, + metainfo=dict(dataset_type='set5', task_name='sisr'), + data_root=data_root + '/Set5', + data_prefix=dict(img='LRbicx3', gt='GTmod12'), + pipeline=val_pipeline)) + +val_evaluator = [ + dict(type='PSNR', crop_border=scale), + dict(type='SSIM', crop_border=scale), +] + +train_cfg = dict( + type='IterBasedTrainLoop', max_iters=500_000, val_interval=5000) +val_cfg = dict(type='ValLoop') + +# optimizer +optim_wrapper = dict( + constructor='DefaultOptimWrapperConstructor', + type='OptimWrapper', + optimizer=dict(type='Adam', lr=2e-4, betas=(0.9, 0.999))) + +# learning policy +param_scheduler = dict( + type='MultiStepLR', + by_epoch=False, + milestones=[250000, 400000, 450000, 475000], + gamma=0.5) diff --git a/configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py b/configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py new file mode 100644 index 0000000000..6f929ec8d7 --- /dev/null +++ b/configs/swinir/swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k.py @@ -0,0 +1,22 @@ +_base_ = ['swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py'] + +experiment_name = 'swinir_x3s64w8d4e60_8xb4-lr2e-4-500k_div2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 3 +img_size = 64 + +# model settings +model = dict( + generator=dict( + img_size=img_size, + depths=[6, 6, 6, 6], + embed_dim=60, + num_heads=[6, 6, 6, 6], + upsampler='pixelshuffledirect')) + +# modify patch size of train_dataloader +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[3]['gt_patch_size'] = img_size * scale diff --git a/configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py b/configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py new file mode 100644 index 0000000000..656046e9d1 --- /dev/null +++ b/configs/swinir/swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k.py @@ -0,0 +1,35 @@ +_base_ = ['swinir_x3s48w8d6e180_8xb4-lr2e-4-500k_div2k.py'] + +experiment_name = 'swinir_x3s64w8d6e180_8xb4-lr2e-4-500k_df2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 3 +img_size = 64 + +# model settings +model = dict(generator=dict(img_size=img_size)) + +# modify patch size of train_pipeline +train_pipeline = _base_.train_pipeline +train_pipeline[3]['gt_patch_size'] = img_size * scale + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=4, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DF2K3450sub_GT.txt', + metainfo=dict(dataset_type='div2k', task_name='sisr'), + data_root=data_root + '/DF2K', + data_prefix=dict( + img='DF2K_train_LR_bicubic/X3_sub', gt='DF2K_train_HR_sub'), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) diff --git a/configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py b/configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py new file mode 100644 index 0000000000..7409802acd --- /dev/null +++ b/configs/swinir/swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py @@ -0,0 +1,133 @@ +_base_ = [ + '../_base_/default_runtime.py', '../_base_/datasets/sisr_x4_test_config.py' +] + +experiment_name = 'swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 4 +img_size = 48 + +# evaluated on Y channels +test_evaluator = _base_.test_evaluator +for evaluator in test_evaluator: + for metric in evaluator: + metric['convert_to'] = 'Y' + +# model settings +model = dict( + type='BaseEditModel', + generator=dict( + type='SwinIRNet', + upscale=scale, + in_chans=3, + img_size=img_size, + window_size=8, + img_range=1.0, + depths=[6, 6, 6, 6, 6, 6], + embed_dim=180, + num_heads=[6, 6, 6, 6, 6, 6], + mlp_ratio=2, + upsampler='pixelshuffle', + resi_connection='1conv'), + pixel_loss=dict(type='L1Loss', loss_weight=1.0, reduction='mean'), + data_preprocessor=dict( + type='EditDataPreprocessor', mean=[0., 0., 0.], std=[255., 255., + 255.])) + +train_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='SetValues', dictionary=dict(scale=scale)), + dict(type='PairedRandomCrop', gt_patch_size=img_size * scale), + dict( + type='Flip', + keys=['img', 'gt'], + flip_ratio=0.5, + direction='horizontal'), + dict( + type='Flip', keys=['img', 'gt'], flip_ratio=0.5, direction='vertical'), + dict(type='RandomTransposeHW', keys=['img', 'gt'], transpose_ratio=0.5), + dict(type='PackEditInputs') +] + +val_pipeline = [ + dict( + type='LoadImageFromFile', + key='img', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict( + type='LoadImageFromFile', + key='gt', + color_type='color', + channel_order='rgb', + imdecode_backend='cv2'), + dict(type='PackEditInputs') +] + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=4, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DIV2K800sub_GT.txt', + metainfo=dict(dataset_type='div2k', task_name='sisr'), + data_root=data_root + '/DIV2K', + data_prefix=dict( + img='DIV2K_train_LR_bicubic/X4_sub', gt='DIV2K_train_HR_sub'), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) + +val_dataloader = dict( + num_workers=4, + persistent_workers=False, + drop_last=False, + sampler=dict(type='DefaultSampler', shuffle=False), + dataset=dict( + type=dataset_type, + metainfo=dict(dataset_type='set5', task_name='sisr'), + data_root=data_root + '/Set5', + data_prefix=dict(img='LRbicx4', gt='GTmod12'), + pipeline=val_pipeline)) + +val_evaluator = [ + dict(type='PSNR', crop_border=scale), + dict(type='SSIM', crop_border=scale), +] + +train_cfg = dict( + type='IterBasedTrainLoop', max_iters=500_000, val_interval=5000) +val_cfg = dict(type='ValLoop') + +# optimizer +optim_wrapper = dict( + constructor='DefaultOptimWrapperConstructor', + type='OptimWrapper', + optimizer=dict(type='Adam', lr=2e-4, betas=(0.9, 0.999))) + +# learning policy +param_scheduler = dict( + type='MultiStepLR', + by_epoch=False, + milestones=[250000, 400000, 450000, 475000], + gamma=0.5) diff --git a/configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py b/configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py new file mode 100644 index 0000000000..09a08ed95a --- /dev/null +++ b/configs/swinir/swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k.py @@ -0,0 +1,22 @@ +_base_ = ['swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py'] + +experiment_name = 'swinir_x4s64w8d4e60_8xb4-lr2e-4-500k_div2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 4 +img_size = 64 + +# model settings +model = dict( + generator=dict( + img_size=img_size, + depths=[6, 6, 6, 6], + embed_dim=60, + num_heads=[6, 6, 6, 6], + upsampler='pixelshuffledirect')) + +# modify patch size of train_dataloader +train_dataloader = _base_.train_dataloader +train_pipeline = train_dataloader['dataset']['pipeline'] +train_pipeline[3]['gt_patch_size'] = img_size * scale diff --git a/configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py b/configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py new file mode 100644 index 0000000000..87cdf5e609 --- /dev/null +++ b/configs/swinir/swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k.py @@ -0,0 +1,35 @@ +_base_ = ['swinir_x4s48w8d6e180_8xb4-lr2e-4-500k_div2k.py'] + +experiment_name = 'swinir_x4s64w8d6e180_8xb4-lr2e-4-500k_df2k' +work_dir = f'./work_dirs/{experiment_name}' +save_dir = './work_dirs/' + +scale = 4 +img_size = 64 + +# model settings +model = dict(generator=dict(img_size=img_size)) + +# modify patch size of train_pipeline +train_pipeline = _base_.train_pipeline +train_pipeline[3]['gt_patch_size'] = img_size * scale + +# dataset settings +dataset_type = 'BasicImageDataset' +data_root = 'data' + +train_dataloader = dict( + num_workers=4, + batch_size=4, + drop_last=True, + persistent_workers=False, + sampler=dict(type='InfiniteSampler', shuffle=True), + dataset=dict( + type=dataset_type, + ann_file='meta_info_DF2K3450sub_GT.txt', + metainfo=dict(dataset_type='div2k', task_name='sisr'), + data_root=data_root + '/DF2K', + data_prefix=dict( + img='DF2K_train_LR_bicubic/X4_sub', gt='DF2K_train_HR_sub'), + filename_tmpl=dict(img='{}', gt='{}'), + pipeline=train_pipeline)) diff --git a/mmedit/datasets/transforms/random_degradations.py b/mmedit/datasets/transforms/random_degradations.py index 65e7fb7849..1c202125aa 100644 --- a/mmedit/datasets/transforms/random_degradations.py +++ b/mmedit/datasets/transforms/random_degradations.py @@ -162,11 +162,13 @@ class RandomJPEGCompression: params (dict): A dictionary specifying the degradation settings. keys (list[str]): A list specifying the keys whose values are modified. + bgr2rgb (str): Whether change channel order. Default: False. """ - def __init__(self, params, keys): + def __init__(self, params, keys, bgr2rgb=False): self.keys = keys self.params = params + self.bgr2rgb = bgr2rgb def _apply_random_compression(self, imgs): is_single_image = False @@ -176,6 +178,7 @@ def _apply_random_compression(self, imgs): # determine initial compression level and the step size quality = self.params['quality'] + color_type = self.params['color_type'] quality_step = self.params.get('quality_step', 0) jpeg_param = round(np.random.uniform(quality[0], quality[1])) @@ -183,8 +186,17 @@ def _apply_random_compression(self, imgs): outputs = [] for img in imgs: encode_param = [int(cv2.IMWRITE_JPEG_QUALITY), jpeg_param] - _, img_encoded = cv2.imencode('.jpg', img * 255., encode_param) - outputs.append(np.float32(cv2.imdecode(img_encoded, 1)) / 255.) + if self.bgr2rgb and color_type == 'color': + img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) + _, img_encoded = cv2.imencode('.jpg', img, encode_param) + + if color_type == 'color': + img_encoded = cv2.imdecode(img_encoded, 1) + if self.bgr2rgb: + img_encoded = cv2.cvtColor(img_encoded, cv2.COLOR_BGR2RGB) + outputs.append(img_encoded) + else: + outputs.append(cv2.imdecode(img_encoded, 0)) # update compression level jpeg_param += np.random.uniform(-quality_step, quality_step) @@ -238,7 +250,7 @@ def _apply_gaussian_noise(self, imgs): Tensor: images applied gaussian noise """ sigma_range = self.params['gaussian_sigma'] - sigma = np.random.uniform(sigma_range[0], sigma_range[1]) / 255. + sigma = np.random.uniform(sigma_range[0], sigma_range[1]) sigma_step = self.params.get('gaussian_sigma_step', 0) @@ -253,9 +265,8 @@ def _apply_gaussian_noise(self, imgs): outputs.append(img + noise) # update noise level - sigma += np.random.uniform(-sigma_step, sigma_step) / 255. - sigma = np.clip(sigma, sigma_range[0] / 255., - sigma_range[1] / 255.) + sigma += np.random.uniform(-sigma_step, sigma_step) + sigma = np.clip(sigma, sigma_range[0], sigma_range[1]) return outputs @@ -274,7 +285,7 @@ def _apply_poisson_noise(self, imgs): if is_gray_noise: noise = cv2.cvtColor(noise[..., [2, 1, 0]], cv2.COLOR_BGR2GRAY) noise = noise[..., np.newaxis] - noise = np.clip((noise * 255.0).round(), 0, 255) / 255. + noise = np.clip((noise).round(), 0, 255) unique_val = 2**np.ceil(np.log2(len(np.unique(noise)))) noise = np.random.poisson(noise * unique_val) / unique_val - noise diff --git a/mmedit/models/editors/__init__.py b/mmedit/models/editors/__init__.py index a59ad82b0a..b1add6f756 100644 --- a/mmedit/models/editors/__init__.py +++ b/mmedit/models/editors/__init__.py @@ -55,6 +55,7 @@ from .stylegan1 import StyleGAN1 from .stylegan2 import StyleGAN2 from .stylegan3 import StyleGAN3, StyleGAN3Generator +from .swinir import SwinIRNet from .tdan import TDAN, TDANNet from .tof import TOFlowVFINet, TOFlowVSRNet, ToFResBlock from .ttsr import LTE, TTSR, SearchTransformer, TTSRDiscriminator, TTSRNet @@ -86,5 +87,5 @@ 'StyleGAN3Generator', 'InstColorization', 'NAFBaseline', 'NAFBaselineLocal', 'NAFNet', 'NAFNetLocal', 'DDIMScheduler', 'DDPMScheduler', 'DenoisingUnet', 'ClipWrapper', 'EG3D', 'Restormer', - 'StableDiffusion' + 'SwinIRNet', 'StableDiffusion' ] diff --git a/mmedit/models/editors/swinir/__init__.py b/mmedit/models/editors/swinir/__init__.py new file mode 100644 index 0000000000..18c102944a --- /dev/null +++ b/mmedit/models/editors/swinir/__init__.py @@ -0,0 +1,4 @@ +# Copyright (c) OpenMMLab. All rights reserved. +from .swinir_net import SwinIRNet + +__all__ = ['SwinIRNet'] diff --git a/mmedit/models/editors/swinir/swinir_modules.py b/mmedit/models/editors/swinir/swinir_modules.py new file mode 100644 index 0000000000..42d6a9c6fc --- /dev/null +++ b/mmedit/models/editors/swinir/swinir_modules.py @@ -0,0 +1,148 @@ +# Copyright (c) OpenMMLab. All rights reserved. +import math + +import torch.nn as nn + +from .swinir_utils import to_2tuple + + +class PatchEmbed(nn.Module): + r""" Image to Patch Embedding + Args: + img_size (int): Image size. Default: 224. + patch_size (int): Patch token size. Default: 4. + in_chans (int): Number of input image channels. Default: 3. + embed_dim (int): Number of linear projection output channels. + Default: 96. + norm_layer (nn.Module, optional): Normalization layer. Default: None + """ + + def __init__(self, + img_size=224, + patch_size=4, + in_chans=3, + embed_dim=96, + norm_layer=None): + super().__init__() + img_size = to_2tuple(img_size) + patch_size = to_2tuple(patch_size) + patches_resolution = [ + img_size[0] // patch_size[0], img_size[1] // patch_size[1] + ] + self.img_size = img_size + self.patch_size = patch_size + self.patches_resolution = patches_resolution + self.num_patches = patches_resolution[0] * patches_resolution[1] + + self.in_chans = in_chans + self.embed_dim = embed_dim + + if norm_layer is not None: + self.norm = norm_layer(embed_dim) + else: + self.norm = None + + def forward(self, x): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, C, Ph, Pw). + + Returns: + Tensor: Forward results. + """ + x = x.flatten(2).transpose(1, 2) # B Ph*Pw C + if self.norm is not None: + x = self.norm(x) + return x + + +class PatchUnEmbed(nn.Module): + r""" Image to Patch Unembedding + Args: + img_size (int): Image size. Default: 224. + patch_size (int): Patch token size. Default: 4. + in_chans (int): Number of input image channels. Default: 3. + embed_dim (int): Number of linear projection output channels. + Default: 96. + norm_layer (nn.Module, optional): Normalization layer. Default: None + """ + + def __init__(self, + img_size=224, + patch_size=4, + in_chans=3, + embed_dim=96, + norm_layer=None): + super().__init__() + img_size = to_2tuple(img_size) + patch_size = to_2tuple(patch_size) + patches_resolution = [ + img_size[0] // patch_size[0], img_size[1] // patch_size[1] + ] + self.img_size = img_size + self.patch_size = patch_size + self.patches_resolution = patches_resolution + self.num_patches = patches_resolution[0] * patches_resolution[1] + + self.in_chans = in_chans + self.embed_dim = embed_dim + + def forward(self, x, x_size): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, L, C). + x_size (tuple[int]): Resolution of input feature. + + Returns: + Tensor: Forward results. + """ + B, HW, C = x.shape + x = x.transpose(1, 2).view(B, self.embed_dim, x_size[0], + x_size[1]) # B Ph*Pw C + return x + + +class Upsample(nn.Sequential): + """Upsample module. + + Args: + scale (int): Scale factor. Supported scales: 2^n and 3. + num_feat (int): Channel number of intermediate features. + """ + + def __init__(self, scale, num_feat): + m = [] + if (scale & (scale - 1)) == 0: # scale = 2^n + for _ in range(int(math.log(scale, 2))): + m.append(nn.Conv2d(num_feat, 4 * num_feat, 3, 1, 1)) + m.append(nn.PixelShuffle(2)) + elif scale == 3: + m.append(nn.Conv2d(num_feat, 9 * num_feat, 3, 1, 1)) + m.append(nn.PixelShuffle(3)) + else: + raise ValueError(f'scale {scale} is not supported. ' + 'Supported scales: 2^n and 3.') + super(Upsample, self).__init__(*m) + + +class UpsampleOneStep(nn.Sequential): + """UpsampleOneStep module (the difference with Upsample is that it always + only has 1conv + 1pixelshuffle) Used in lightweight SR to save parameters. + + Args: + scale (int): Scale factor. Supported scales: 2^n and 3. + num_feat (int): Channel number of intermediate features. + num_out_ch (int): Channel number for PixelShuffle. + input_resolution (tuple[int], optional): Input resolution. + Default: None + """ + + def __init__(self, scale, num_feat, num_out_ch, input_resolution=None): + self.num_feat = num_feat + self.input_resolution = input_resolution + m = [] + m.append(nn.Conv2d(num_feat, (scale**2) * num_out_ch, 3, 1, 1)) + m.append(nn.PixelShuffle(scale)) + super(UpsampleOneStep, self).__init__(*m) diff --git a/mmedit/models/editors/swinir/swinir_net.py b/mmedit/models/editors/swinir/swinir_net.py new file mode 100644 index 0000000000..859cfa634b --- /dev/null +++ b/mmedit/models/editors/swinir/swinir_net.py @@ -0,0 +1,307 @@ +# Copyright (c) OpenMMLab. All rights reserved. +import torch +import torch.nn as nn +import torch.nn.functional as F +from mmengine.model import BaseModule +from mmengine.model.weight_init import trunc_normal_ + +from mmedit.registry import MODELS +from .swinir_modules import PatchEmbed, PatchUnEmbed, Upsample, UpsampleOneStep +from .swinir_rstb import RSTB + + +@MODELS.register_module() +class SwinIRNet(BaseModule): + r""" SwinIR + A PyTorch impl of: `SwinIR: Image Restoration Using Swin Transformer`, + based on Swin Transformer. + Ref repo: https://github.com/JingyunLiang/SwinIR + + Args: + img_size (int | tuple(int)): Input image size. Default 64 + patch_size (int | tuple(int)): Patch size. Default: 1 + in_chans (int): Number of input image channels. Default: 3 + embed_dim (int): Patch embedding dimension. Default: 96 + depths (tuple(int)): Depth of each Swin Transformer layer. + Default: [6, 6, 6, 6] + num_heads (tuple(int)): Number of attention heads in different layers. + Default: [6, 6, 6, 6] + window_size (int): Window size. Default: 7 + mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4 + qkv_bias (bool): If True, add a learnable bias to query, key, value. + Default: True + qk_scale (float): Override default qk scale of head_dim ** -0.5 if set. + Default: None + drop_rate (float): Dropout rate. Default: 0 + attn_drop_rate (float): Attention dropout rate. Default: 0 + drop_path_rate (float): Stochastic depth rate. Default: 0.1 + norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm. + ape (bool): If True, add absolute position embedding to the + patch embedding. Default: False + patch_norm (bool): If True, add normalization after patch embedding. + Default: True + use_checkpoint (bool): Whether to use checkpointing to save memory. + Default: False + upscale (int): Upscale factor. 2/3/4/8 for image SR, 1 for denoising + and compress artifact reduction. Default: 2 + img_range (float): Image range. 1. or 255. Default: 1.0 + upsampler (string, optional): The reconstruction module. + 'pixelshuffle' / 'pixelshuffledirect' /'nearest+conv'/None. + Default: '' + resi_connection (string): The convolutional block before residual + connection. '1conv'/'3conv'. Default: '1conv' + """ + + def __init__(self, + img_size=64, + patch_size=1, + in_chans=3, + embed_dim=96, + depths=[6, 6, 6, 6], + num_heads=[6, 6, 6, 6], + window_size=7, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + drop_rate=0., + attn_drop_rate=0., + drop_path_rate=0.1, + norm_layer=nn.LayerNorm, + ape=False, + patch_norm=True, + use_checkpoint=False, + upscale=2, + img_range=1., + upsampler='', + resi_connection='1conv', + **kwargs): + super(SwinIRNet, self).__init__() + num_in_ch = in_chans + num_out_ch = in_chans + num_feat = 64 + self.img_range = img_range + if in_chans == 3: + rgb_mean = (0.4488, 0.4371, 0.4040) + self.mean = torch.Tensor(rgb_mean).view(1, 3, 1, 1) + else: + self.mean = torch.zeros(1, 1, 1, 1) + self.upscale = upscale + self.upsampler = upsampler + self.window_size = window_size + + # 1, shallow feature extraction + self.conv_first = nn.Conv2d(num_in_ch, embed_dim, 3, 1, 1) + + # 2, deep feature extraction + self.num_layers = len(depths) + self.embed_dim = embed_dim + self.ape = ape + self.patch_norm = patch_norm + self.num_features = embed_dim + self.mlp_ratio = mlp_ratio + + # split image into non-overlapping patches + self.patch_embed = PatchEmbed( + img_size=img_size, + patch_size=patch_size, + in_chans=embed_dim, + embed_dim=embed_dim, + norm_layer=norm_layer if self.patch_norm else None) + num_patches = self.patch_embed.num_patches + patches_resolution = self.patch_embed.patches_resolution + self.patches_resolution = patches_resolution + + # merge non-overlapping patches into image + self.patch_unembed = PatchUnEmbed( + img_size=img_size, + patch_size=patch_size, + in_chans=embed_dim, + embed_dim=embed_dim, + norm_layer=norm_layer if self.patch_norm else None) + + # absolute position embedding + if self.ape: + self.absolute_pos_embed = nn.Parameter( + torch.zeros(1, num_patches, embed_dim)) + trunc_normal_(self.absolute_pos_embed, std=.02) + + self.pos_drop = nn.Dropout(p=drop_rate) + + # stochastic depth decay rule + dpr = [ + x.item() for x in torch.linspace(0, drop_path_rate, sum(depths)) + ] + + # build Residual Swin Transformer blocks (RSTB) + self.layers = nn.ModuleList() + for i_layer in range(self.num_layers): + layer = RSTB( + dim=embed_dim, + input_resolution=(patches_resolution[0], + patches_resolution[1]), + depth=depths[i_layer], + num_heads=num_heads[i_layer], + window_size=window_size, + mlp_ratio=self.mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop_rate, + attn_drop=attn_drop_rate, + drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], + norm_layer=norm_layer, + downsample=None, + use_checkpoint=use_checkpoint, + img_size=img_size, + patch_size=patch_size, + resi_connection=resi_connection) + self.layers.append(layer) + self.norm = norm_layer(self.num_features) + + # build the last conv layer in deep feature extraction + if resi_connection == '1conv': + self.conv_after_body = nn.Conv2d(embed_dim, embed_dim, 3, 1, 1) + elif resi_connection == '3conv': + # to save parameters and memory + self.conv_after_body = nn.Sequential( + nn.Conv2d(embed_dim, embed_dim // 4, 3, 1, 1), + nn.LeakyReLU(negative_slope=0.2, inplace=True), + nn.Conv2d(embed_dim // 4, embed_dim // 4, 1, 1, 0), + nn.LeakyReLU(negative_slope=0.2, inplace=True), + nn.Conv2d(embed_dim // 4, embed_dim, 3, 1, 1)) + + # 3, high quality image reconstruction + if self.upsampler == 'pixelshuffle': + # for classical SR + self.conv_before_upsample = nn.Sequential( + nn.Conv2d(embed_dim, num_feat, 3, 1, 1), + nn.LeakyReLU(inplace=True)) + self.upsample = Upsample(upscale, num_feat) + self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) + elif self.upsampler == 'pixelshuffledirect': + # for lightweight SR (to save parameters) + self.upsample = UpsampleOneStep(upscale, embed_dim, num_out_ch, + (patches_resolution[0], + patches_resolution[1])) + elif self.upsampler == 'nearest+conv': + # for real-world SR (less artifacts) + self.conv_before_upsample = nn.Sequential( + nn.Conv2d(embed_dim, num_feat, 3, 1, 1), + nn.LeakyReLU(inplace=True)) + self.conv_up1 = nn.Conv2d(num_feat, num_feat, 3, 1, 1) + if self.upscale == 4: + self.conv_up2 = nn.Conv2d(num_feat, num_feat, 3, 1, 1) + self.conv_hr = nn.Conv2d(num_feat, num_feat, 3, 1, 1) + self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1) + self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True) + else: + # for image denoising and JPEG compression artifact reduction + self.conv_last = nn.Conv2d(embed_dim, num_out_ch, 3, 1, 1) + + self.apply(self._init_weights) + + def _init_weights(self, m): + if isinstance(m, nn.Linear): + trunc_normal_(m.weight, std=.02) + if isinstance(m, nn.Linear) and m.bias is not None: + nn.init.constant_(m.bias, 0) + elif isinstance(m, nn.LayerNorm): + nn.init.constant_(m.bias, 0) + nn.init.constant_(m.weight, 1.0) + + @torch.jit.ignore + def no_weight_decay(self): + return {'absolute_pos_embed'} + + @torch.jit.ignore + def no_weight_decay_keywords(self): + return {'relative_position_bias_table'} + + def check_image_size(self, x): + """Check image size and pad images so that it has enough dimension do + window size. + + args: + x: input tensor image with (B, C, H, W) shape. + """ + _, _, h, w = x.size() + mod_pad_h = (self.window_size - + h % self.window_size) % self.window_size + mod_pad_w = (self.window_size - + w % self.window_size) % self.window_size + x = F.pad(x, (0, mod_pad_w, 0, mod_pad_h), 'reflect') + return x + + def forward_features(self, x): + """Forward function of Deep Feature Extraction. + + Args: + x (Tensor): Input tensor with shape (B, C, H, W). + + Returns: + Tensor: Forward results. + """ + x_size = (x.shape[2], x.shape[3]) + x = self.patch_embed(x) + if self.ape: + x = x + self.absolute_pos_embed + x = self.pos_drop(x) + + for layer in self.layers: + x = layer(x, x_size) + + x = self.norm(x) # B L C + x = self.patch_unembed(x, x_size) + + return x + + def forward(self, x): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, C, H, W). + + Returns: + Tensor: Forward results. + """ + H, W = x.shape[2:] + x = self.check_image_size(x) + + self.mean = self.mean.type_as(x) + x = (x - self.mean) * self.img_range + + if self.upsampler == 'pixelshuffle': + # for classical SR + x = self.conv_first(x) + x = self.conv_after_body(self.forward_features(x)) + x + x = self.conv_before_upsample(x) + x = self.conv_last(self.upsample(x)) + elif self.upsampler == 'pixelshuffledirect': + # for lightweight SR + x = self.conv_first(x) + x = self.conv_after_body(self.forward_features(x)) + x + x = self.upsample(x) + elif self.upsampler == 'nearest+conv': + # for real-world SR + x = self.conv_first(x) + x = self.conv_after_body(self.forward_features(x)) + x + x = self.conv_before_upsample(x) + x = self.lrelu( + self.conv_up1( + torch.nn.functional.interpolate( + x, scale_factor=2, mode='nearest'))) + if self.upscale == 4: + x = self.lrelu( + self.conv_up2( + torch.nn.functional.interpolate( + x, scale_factor=2, mode='nearest'))) + x = self.conv_last(self.lrelu(self.conv_hr(x))) + else: + # for image denoising and JPEG compression artifact reduction + x_first = self.conv_first(x) + res = self.conv_after_body( + self.forward_features(x_first)) + x_first + x = x + self.conv_last(res) + + x = x / self.img_range + self.mean + + return x[:, :, :H * self.upscale, :W * self.upscale] diff --git a/mmedit/models/editors/swinir/swinir_rstb.py b/mmedit/models/editors/swinir/swinir_rstb.py new file mode 100644 index 0000000000..48b0b56d3b --- /dev/null +++ b/mmedit/models/editors/swinir/swinir_rstb.py @@ -0,0 +1,583 @@ +# Copyright (c) OpenMMLab. All rights reserved. + +import torch +import torch.nn as nn +import torch.utils.checkpoint as checkpoint +from mmengine.model.weight_init import trunc_normal_ + +from .swinir_modules import PatchEmbed, PatchUnEmbed +from .swinir_utils import (drop_path, to_2tuple, window_partition, + window_reverse) + + +class DropPath(nn.Module): + """Drop paths (Stochastic Depth) per sample (when applied in main path of + residual blocks).""" + + def __init__(self, drop_prob: float = 0., scale_by_keep: bool = True): + super(DropPath, self).__init__() + self.drop_prob = drop_prob + self.scale_by_keep = scale_by_keep + + def forward(self, x): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, L, C). + + Returns: + Tensor: Forward results. + """ + return drop_path(x, self.drop_prob, self.training, self.scale_by_keep) + + def extra_repr(self): + return f'drop_prob={round(self.drop_prob, 3):0.3f}' + + +class Mlp(nn.Module): + """Multilayer Perceptron layer. + + Args: + in_features (int): Number of input channels. + hidden_features (int | None, optional): Number of hidden layer + channels. Default: None + out_features (int | None, optional): Number of output channels. + Default: None + act_layer (nn.Module, optional): Activation layer. Default: nn.GELU + drop (float, optional): Dropout ratio of attention weight. Default: 0.0 + """ + + def __init__(self, + in_features, + hidden_features=None, + out_features=None, + act_layer=nn.GELU, + drop=0.): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Linear(in_features, hidden_features) + self.act = act_layer() + self.fc2 = nn.Linear(hidden_features, out_features) + self.drop = nn.Dropout(drop) + + def forward(self, x): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, L, C). + + Returns: + Tensor: Forward results. + """ + x = self.fc1(x) + x = self.act(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +class WindowAttention(nn.Module): + r""" Window based multi-head self attention (W-MSA) + module with relative position bias. + It supports both of shifted and non-shifted window. + Args: + dim (int): Number of input channels. + window_size (tuple[int]): The height and width of the window. + num_heads (int): Number of attention heads. + qkv_bias (bool, optional): If True, add a learnable bias to + query, key, value. Default: True + qk_scale (float | None, optional): Override default qk scale + of head_dim ** -0.5 if set + attn_drop (float, optional): Dropout ratio of attention weight. + Default: 0.0 + proj_drop (float, optional): Dropout ratio of output. Default: 0.0 + """ + + def __init__(self, + dim, + window_size, + num_heads, + qkv_bias=True, + qk_scale=None, + attn_drop=0., + proj_drop=0.): + + super().__init__() + self.dim = dim + self.window_size = window_size # Wh, Ww + self.num_heads = num_heads + head_dim = dim // num_heads + self.scale = qk_scale or head_dim**-0.5 + + # define a parameter table of relative position bias + # 2*Wh-1 * 2*Ww-1, nH + self.relative_position_bias_table = nn.Parameter( + torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), + num_heads)) + + # get pair-wise relative position index + # for each token inside the window + coords_h = torch.arange(self.window_size[0]) + coords_w = torch.arange(self.window_size[1]) + coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww + coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww + relative_coords = \ + coords_flatten[:, :, None] - coords_flatten[:, None, :] + # Wh*Ww, Wh*Ww, 2 + relative_coords = relative_coords.permute(1, 2, 0).contiguous() + # shift to start from 0 + relative_coords[:, :, 0] += self.window_size[0] - 1 + relative_coords[:, :, 1] += self.window_size[1] - 1 + relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1 + relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww + self.register_buffer('relative_position_index', + relative_position_index) + + self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) + self.attn_drop = nn.Dropout(attn_drop) + self.proj = nn.Linear(dim, dim) + + self.proj_drop = nn.Dropout(proj_drop) + + trunc_normal_(self.relative_position_bias_table, std=.02) + self.softmax = nn.Softmax(dim=-1) + + def forward(self, x, mask=None): + """ + Args: + x: input features with shape of (num_windows*B, N, C) + mask: (0/-inf) mask with shape of + (num_windows, Wh*Ww, Wh*Ww) or None + """ + B_, N, C = x.shape + qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, + C // self.num_heads).permute(2, 0, 3, 1, 4) + # make torchscript happy (cannot use tensor as tuple) + q, k, v = qkv[0], qkv[1], qkv[2] + + q = q * self.scale + attn = (q @ k.transpose(-2, -1)) + + relative_position_bias = self.relative_position_bias_table[ + self.relative_position_index.view(-1)].view( + self.window_size[0] * self.window_size[1], + self.window_size[0] * self.window_size[1], + -1) # Wh*Ww,Wh*Ww,nH + relative_position_bias = relative_position_bias.permute( + 2, 0, 1).contiguous() # nH, Wh*Ww, Wh*Ww + attn = attn + relative_position_bias.unsqueeze(0) + + if mask is not None: + nW = mask.shape[0] + attn = attn.view(B_ // nW, nW, self.num_heads, N, + N) + mask.unsqueeze(1).unsqueeze(0) + attn = attn.view(-1, self.num_heads, N, N) + attn = self.softmax(attn) + else: + attn = self.softmax(attn) + + attn = self.attn_drop(attn) + + x = (attn @ v).transpose(1, 2).reshape(B_, N, C) + x = self.proj(x) + x = self.proj_drop(x) + return x + + def extra_repr(self) -> str: + return f'dim={self.dim}, window_size={self.window_size}, ' \ + f'num_heads={self.num_heads}' + + +class SwinTransformerBlock(nn.Module): + r""" Swin Transformer Block. + Args: + dim (int): Number of input channels. + input_resolution (tuple[int]): Input resolution. + num_heads (int): Number of attention heads. + window_size (int): Window size. + shift_size (int): Shift size for SW-MSA. + mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. + qkv_bias (bool, optional): If True, add a learnable bias + to query, key, value. Default: True + qk_scale (float | None, optional): Override default qk + scale of head_dim ** -0.5 if set. + drop (float, optional): Dropout rate. Default: 0.0 + attn_drop (float, optional): Attention dropout rate. Default: 0.0 + drop_path (float, optional): Stochastic depth rate. Default: 0.0 + act_layer (nn.Module, optional): Activation layer. Default: nn.GELU + norm_layer (nn.Module, optional): Normalization layer. + Default: nn.LayerNorm + """ + + def __init__(self, + dim, + input_resolution, + num_heads, + window_size=7, + shift_size=0, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + act_layer=nn.GELU, + norm_layer=nn.LayerNorm): + super().__init__() + self.dim = dim + self.input_resolution = input_resolution + self.num_heads = num_heads + self.window_size = window_size + self.shift_size = shift_size + self.mlp_ratio = mlp_ratio + if min(self.input_resolution) <= self.window_size: + # if window size is larger than input resolution, + # we don't partition windows + self.shift_size = 0 + self.window_size = min(self.input_resolution) + assert 0 <= self.shift_size < self.window_size, \ + 'shift_size must in 0-window_size' + + self.norm1 = norm_layer(dim) + self.attn = WindowAttention( + dim, + window_size=to_2tuple(self.window_size), + num_heads=num_heads, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + attn_drop=attn_drop, + proj_drop=drop) + + self.drop_path = DropPath( + drop_path) if drop_path > 0. else nn.Identity() + self.norm2 = norm_layer(dim) + mlp_hidden_dim = int(dim * mlp_ratio) + self.mlp = Mlp( + in_features=dim, + hidden_features=mlp_hidden_dim, + act_layer=act_layer, + drop=drop) + + if self.shift_size > 0: + attn_mask = self.calculate_mask(self.input_resolution) + else: + attn_mask = None + + self.register_buffer('attn_mask', attn_mask) + + def calculate_mask(self, x_size): + # calculate attention mask for SW-MSA + """Calculate attention mask for SW-MSA. + + Args: + x_size (tuple[int]): Resolution of input feature. + + Returns: + Tensor: Attention mask + """ + H, W = x_size + img_mask = torch.zeros((1, H, W, 1)) # 1 H W 1 + h_slices = (slice(0, -self.window_size), + slice(-self.window_size, + -self.shift_size), slice(-self.shift_size, None)) + w_slices = (slice(0, -self.window_size), + slice(-self.window_size, + -self.shift_size), slice(-self.shift_size, None)) + cnt = 0 + for h in h_slices: + for w in w_slices: + img_mask[:, h, w, :] = cnt + cnt += 1 + + mask_windows = window_partition( + img_mask, self.window_size) # nW, window_size, window_size, 1 + mask_windows = mask_windows.view(-1, + self.window_size * self.window_size) + attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) + attn_mask = attn_mask.masked_fill(attn_mask != 0, + float(-100.0)).masked_fill( + attn_mask == 0, float(0.0)) + + return attn_mask + + def forward(self, x, x_size): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, L, C). + x_size (tuple[int]): Resolution of input feature. + + Returns: + Tensor: Forward results. + """ + H, W = x_size + B, L, C = x.shape + # assert L == H * W, "input feature has wrong size" + + shortcut = x + x = self.norm1(x) + x = x.view(B, H, W, C) + + # cyclic shift + if self.shift_size > 0: + shifted_x = torch.roll( + x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)) + else: + shifted_x = x + + # partition windows + x_windows = window_partition( + shifted_x, self.window_size) # nW*B, window_size, window_size, C + x_windows = x_windows.view(-1, self.window_size * self.window_size, + C) # nW*B, window_size*window_size, C + + # W-MSA/SW-MSA (to be compatible for testing on images + # whose shapes are the multiple of window size + if self.input_resolution == x_size: + attn_windows = self.attn( + x_windows, + mask=self.attn_mask) # nW*B, window_size*window_size, C + else: + attn_windows = self.attn( + x_windows, mask=self.calculate_mask(x_size).to(x.device)) + + # merge windows + attn_windows = attn_windows.view(-1, self.window_size, + self.window_size, C) + shifted_x = window_reverse(attn_windows, self.window_size, H, + W) # B H' W' C + + # reverse cyclic shift + if self.shift_size > 0: + x = torch.roll( + shifted_x, + shifts=(self.shift_size, self.shift_size), + dims=(1, 2)) + else: + x = shifted_x + x = x.view(B, H * W, C) + + # FFN + x = shortcut + self.drop_path(x) + x = x + self.drop_path(self.mlp(self.norm2(x))) + + return x + + def extra_repr(self) -> str: + return f'dim={self.dim}, ' \ + f'input_resolution={self.input_resolution}, ' \ + f'num_heads={self.num_heads}, ' \ + f'window_size={self.window_size}, ' \ + f'shift_size={self.shift_size}, ' \ + f'mlp_ratio={self.mlp_ratio}' + + +class BasicLayer(nn.Module): + """A basic Swin Transformer layer for one stage. + + Args: + dim (int): Number of input channels. + input_resolution (tuple[int]): Input resolution. + depth (int): Number of blocks. + num_heads (int): Number of attention heads. + window_size (int): Local window size. + mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. + qkv_bias (bool, optional): If True, add a learnable bias + to query, key, value. Default: True + qk_scale (float | None, optional): Override default qk + scale of head_dim ** -0.5 if set. + drop (float, optional): Dropout rate. Default: 0.0 + attn_drop (float, optional): Attention dropout rate. Default: 0.0 + drop_path (float | tuple[float], optional): Stochastic depth rate. + Default: 0.0 + norm_layer (nn.Module, optional): Normalization layer. + Default: nn.LayerNorm + downsample (nn.Module | None, optional): Downsample layer at the + end of the layer. Default: None + use_checkpoint (bool): Whether to use checkpointing to save memory. + Default: False. + """ + + def __init__(self, + dim, + input_resolution, + depth, + num_heads, + window_size, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + norm_layer=nn.LayerNorm, + downsample=None, + use_checkpoint=False): + + super().__init__() + self.dim = dim + self.input_resolution = input_resolution + self.depth = depth + self.use_checkpoint = use_checkpoint + + # build blocks + self.blocks = nn.ModuleList([ + SwinTransformerBlock( + dim=dim, + input_resolution=input_resolution, + num_heads=num_heads, + window_size=window_size, + shift_size=0 if (i % 2 == 0) else window_size // 2, + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop, + attn_drop=attn_drop, + drop_path=drop_path[i] + if isinstance(drop_path, list) else drop_path, + norm_layer=norm_layer) for i in range(depth) + ]) + + # patch merging layer + if downsample is not None: + self.downsample = downsample( + input_resolution, dim=dim, norm_layer=norm_layer) + else: + self.downsample = None + + def forward(self, x, x_size): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, L, C). + x_size (tuple[int]): Resolution of input feature. + + Returns: + Tensor: Forward results. + """ + for blk in self.blocks: + if self.use_checkpoint: + x = checkpoint.checkpoint(blk, x, x_size) + else: + x = blk(x, x_size) + if self.downsample is not None: + x = self.downsample(x) + return x + + def extra_repr(self) -> str: + return f'dim={self.dim}, ' \ + f'input_resolution={self.input_resolution}, ' \ + f'depth={self.depth}' + + +class RSTB(nn.Module): + """Residual Swin Transformer Block (RSTB). + + Args: + dim (int): Number of input channels. + input_resolution (tuple[int]): Input resolution. + depth (int): Number of blocks. + num_heads (int): Number of attention heads. + window_size (int): Local window size. + mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. + Default: 4.0 + qkv_bias (bool, optional): If True, add a learnable bias to + query, key, value. Default: True + qk_scale (float | None, optional): Override default qk scale + of head_dim ** -0.5 if set. + drop (float, optional): Dropout rate. Default: 0.0 + attn_drop (float, optional): Attention dropout rate. Default: 0.0 + drop_path (float | tuple[float], optional): Stochastic depth rate. + Default: 0.0 + norm_layer (nn.Module, optional): Normalization layer. + Default: nn.LayerNorm + downsample (nn.Module | None, optional): Downsample layer at the + end of the layer. Default: None + use_checkpoint (bool): Whether to use checkpointing to save memory. + Default: False. + img_size (int): Input image size. Default: 224 + patch_size (int): Patch size. Default: 4 + resi_connection (string): The convolutional block before + residual connection. Default: '1conv' + """ + + def __init__(self, + dim, + input_resolution, + depth, + num_heads, + window_size, + mlp_ratio=4., + qkv_bias=True, + qk_scale=None, + drop=0., + attn_drop=0., + drop_path=0., + norm_layer=nn.LayerNorm, + downsample=None, + use_checkpoint=False, + img_size=224, + patch_size=4, + resi_connection='1conv'): + super(RSTB, self).__init__() + + self.dim = dim + self.input_resolution = input_resolution + + self.residual_group = BasicLayer( + dim=dim, + input_resolution=input_resolution, + depth=depth, + num_heads=num_heads, + window_size=window_size, + mlp_ratio=mlp_ratio, + qkv_bias=qkv_bias, + qk_scale=qk_scale, + drop=drop, + attn_drop=attn_drop, + drop_path=drop_path, + norm_layer=norm_layer, + downsample=downsample, + use_checkpoint=use_checkpoint) + + if resi_connection == '1conv': + self.conv = nn.Conv2d(dim, dim, 3, 1, 1) + elif resi_connection == '3conv': + # to save parameters and memory + self.conv = nn.Sequential( + nn.Conv2d(dim, dim // 4, 3, 1, 1), + nn.LeakyReLU(negative_slope=0.2, inplace=True), + nn.Conv2d(dim // 4, dim // 4, 1, 1, 0), + nn.LeakyReLU(negative_slope=0.2, inplace=True), + nn.Conv2d(dim // 4, dim, 3, 1, 1)) + + self.patch_embed = PatchEmbed( + img_size=img_size, + patch_size=patch_size, + in_chans=0, + embed_dim=dim, + norm_layer=None) + + self.patch_unembed = PatchUnEmbed( + img_size=img_size, + patch_size=patch_size, + in_chans=0, + embed_dim=dim, + norm_layer=None) + + def forward(self, x, x_size): + """Forward function. + + Args: + x (Tensor): Input tensor with shape (B, L, C). + x_size (tuple[int]): Resolution of input feature. + + Returns: + Tensor: Forward results. + """ + return self.patch_embed( + self.conv( + self.patch_unembed(self.residual_group(x, x_size), + x_size))) + x diff --git a/mmedit/models/editors/swinir/swinir_utils.py b/mmedit/models/editors/swinir/swinir_utils.py new file mode 100644 index 0000000000..d0f6e69051 --- /dev/null +++ b/mmedit/models/editors/swinir/swinir_utils.py @@ -0,0 +1,84 @@ +# Copyright (c) OpenMMLab. All rights reserved. +import collections.abc +from itertools import repeat + + +# From PyTorch internals +def _ntuple(n): + """A `to_tuple` function generator. It returns a function, this function + will repeat the input to a tuple of length ``n`` if the input is not an + Iterable object, otherwise, return the input directly. + + Args: + n (int): The number of the target length. + """ + + def parse(x): + if isinstance(x, collections.abc.Iterable) and not isinstance(x, str): + return x + return tuple(repeat(x, n)) + + return parse + + +to_2tuple = _ntuple(2) + + +def drop_path(x, + drop_prob: float = 0., + training: bool = False, + scale_by_keep: bool = True): + """Drop paths (Stochastic Depth) per sample (when applied in main path of + residual blocks). + + This is the same as the DropConnect impl I created for + EfficientNet, etc networks, however, the original name is misleading + as 'Drop Connect' is a different form of dropout in a separate paper... + See discussion: + https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 + I've opted for changing the layer and argument names to 'drop path' + rather than mix DropConnect as a layer name and use + 'survival rate' as the argument. + """ + if drop_prob == 0. or not training: + return x + keep_prob = 1 - drop_prob + # work with diff dim tensors, not just 2D ConvNets + shape = (x.shape[0], ) + (1, ) * (x.ndim - 1) + random_tensor = x.new_empty(shape).bernoulli_(keep_prob) + if keep_prob > 0.0 and scale_by_keep: + random_tensor.div_(keep_prob) + return x * random_tensor + + +def window_partition(x, window_size): + """ + Args: + x: (B, H, W, C) + window_size (int): window size + Returns: + windows: (num_windows*B, window_size, window_size, C) + """ + B, H, W, C = x.shape + x = x.view(B, H // window_size, window_size, W // window_size, window_size, + C) + windows = x.permute(0, 1, 3, 2, 4, + 5).contiguous().view(-1, window_size, window_size, C) + return windows + + +def window_reverse(windows, window_size, H, W): + """ + Args: + windows: (num_windows*B, window_size, window_size, C) + window_size (int): Window size + H (int): Height of image + W (int): Width of image + Returns: + x: (B, H, W, C) + """ + B = int(windows.shape[0] / (H * W / window_size / window_size)) + x = windows.view(B, H // window_size, W // window_size, window_size, + window_size, -1) + x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1) + return x diff --git a/model-index.yml b/model-index.yml index bedf04aac2..f1de04f77e 100644 --- a/model-index.yml +++ b/model-index.yml @@ -44,6 +44,7 @@ Import: - configs/styleganv1/metafile.yml - configs/styleganv2/metafile.yml - configs/styleganv3/metafile.yml +- configs/swinir/metafile.yml - configs/tdan/metafile.yml - configs/tof/metafile.yml - configs/ttsr/metafile.yml diff --git a/tests/test_datasets/test_transforms/test_random_degradations.py b/tests/test_datasets/test_transforms/test_random_degradations.py index 2008033322..12546ac5eb 100644 --- a/tests/test_datasets/test_transforms/test_random_degradations.py +++ b/tests/test_datasets/test_transforms/test_random_degradations.py @@ -9,7 +9,7 @@ def test_random_noise(): results = {} - results['lq'] = np.ones((8, 8, 3)).astype(np.float32) + results['lq'] = np.ones((8, 8, 3)).astype(np.uint8) # Gaussian noise model = RandomNoise( @@ -49,17 +49,21 @@ def test_random_noise(): def test_random_jpeg_compression(): results = {} - results['lq'] = np.ones((8, 8, 3)).astype(np.float32) + results['lq'] = np.ones((8, 8, 3)).astype(np.uint8) - model = RandomJPEGCompression(params=dict(quality=[5, 50]), keys=['lq']) + model = RandomJPEGCompression( + params=dict(quality=[5, 50], color_type='color'), keys=['lq']) results = model(results) assert results['lq'].shape == (8, 8, 3) # skip degradations with prob < 1 - params = dict(quality=[5, 50], prob=0) + params = dict(quality=[5, 50], color_type='color', prob=0) model = RandomJPEGCompression(params=params, keys=['lq']) assert model(results) == results + model = RandomJPEGCompression(params=params, keys=['lq'], bgr2rgb=True) + assert model(results)['lq'].shape == results['lq'].shape + assert repr(model) == model.__class__.__name__ + f'(params={params}, ' \ + "keys=['lq'])" @@ -334,7 +338,7 @@ def test_random_blur(): def test_degradations_with_shuffle(): results = {} - results['lq'] = np.ones((8, 8, 3)).astype(np.float32) + results['lq'] = np.ones((8, 8, 3)).astype(np.uint8) # shuffle all model = DegradationsWithShuffle( @@ -360,10 +364,10 @@ def test_degradations_with_shuffle(): [ dict( type='RandomJPEGCompression', - params=dict(quality=[5, 10])), + params=dict(quality=[5, 10], color_type='color')), dict( type='RandomJPEGCompression', - params=dict(quality=[15, 20])) + params=dict(quality=[15, 20], color_type='color')) ] ], keys=['lq'], @@ -391,8 +395,12 @@ def test_degradations_with_shuffle(): resize_prob=[1 / 3., 1 / 3., 1 / 3.], target_size=(16, 16))), [ - dict(type='RandomJPEGCompression', params=dict(quality=[5, 10])), - dict(type='RandomJPEGCompression', params=dict(quality=[15, 20])) + dict( + type='RandomJPEGCompression', + params=dict(quality=[5, 10], color_type='color')), + dict( + type='RandomJPEGCompression', + params=dict(quality=[15, 20], color_type='color')) ] ] model = DegradationsWithShuffle( diff --git a/tests/test_models/test_editors/test_swinir/test_swinir_modules.py b/tests/test_models/test_editors/test_swinir/test_swinir_modules.py new file mode 100644 index 0000000000..3f02556236 --- /dev/null +++ b/tests/test_models/test_editors/test_swinir/test_swinir_modules.py @@ -0,0 +1,84 @@ +# Copyright (c) OpenMMLab. All rights reserved. +import platform + +import pytest +import torch + +from mmedit.models.editors.swinir.swinir_modules import (PatchEmbed, + PatchUnEmbed, + Upsample, + UpsampleOneStep) + + +@pytest.mark.skipif( + 'win' in platform.system().lower() and 'cu' in torch.__version__, + reason='skip on windows-cuda due to limited RAM.') +def test_patchEmbed(): + + net = PatchEmbed( + img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None) + + img = torch.randn(1, 3, 4, 4) + output = net(img) + assert output.shape == (1, 16, 3) + + if torch.cuda.is_available(): + net = net.cuda() + output = net(img.cuda()) + assert output.shape == (1, 16, 3) + + +@pytest.mark.skipif( + 'win' in platform.system().lower() and 'cu' in torch.__version__, + reason='skip on windows-cuda due to limited RAM.') +def test_patchUnEmbed(): + + net = PatchUnEmbed( + img_size=16, patch_size=4, in_chans=3, embed_dim=3, norm_layer=None) + + img = torch.randn(1, 64, 3) + output = net(img, (8, 8)) + assert output.shape == (1, 3, 8, 8) + + if torch.cuda.is_available(): + net = net.cuda() + output = net(img.cuda(), (8, 8)) + assert output.shape == (1, 3, 8, 8) + + +@pytest.mark.skipif( + 'win' in platform.system().lower() and 'cu' in torch.__version__, + reason='skip on windows-cuda due to limited RAM.') +def test_upsample(): + + net = Upsample(scale=2, num_feat=3) + + img = torch.randn(1, 3, 8, 8) + output = net(img) + assert output.shape == (1, 3, 16, 16) + + if torch.cuda.is_available(): + net = net.cuda() + output = net(img.cuda()) + assert output.shape == (1, 3, 16, 16) + + +@pytest.mark.skipif( + 'win' in platform.system().lower() and 'cu' in torch.__version__, + reason='skip on windows-cuda due to limited RAM.') +def test_upsampleOneStep(): + + net = UpsampleOneStep( + scale=2, + num_feat=3, + num_out_ch=4, + ) + + img = torch.randn(1, 3, 8, 8) + output = net(img) + assert output.shape == (1, 4, 16, 16) + + if torch.cuda.is_available(): + net = net.cuda() + output = net(img.cuda()) + assert output.shape == (1, 4, 16, 16) diff --git a/tests/test_models/test_editors/test_swinir/test_swinir_net.py b/tests/test_models/test_editors/test_swinir/test_swinir_net.py new file mode 100644 index 0000000000..b2944baa89 --- /dev/null +++ b/tests/test_models/test_editors/test_swinir/test_swinir_net.py @@ -0,0 +1,108 @@ +# Copyright (c) OpenMMLab. All rights reserved. +import platform + +import pytest +import torch + +from mmedit.models.editors import SwinIRNet + + +def test_swinir_cpu(): + """Test SwinIRNet.""" + + # x2 model classical SR + net = SwinIRNet( + upscale=2, + in_channels=3, + img_size=48, + window_size=8, + img_range=1.0, + depths=[6], + embed_dim=60, + num_heads=[6], + mlp_ratio=2, + upsampler='pixelshuffledirect', + resi_connection='3conv') + img = torch.rand(1, 3, 16, 16) + output = net(img) + assert isinstance(output, torch.Tensor) + assert output.shape == (1, 3, 32, 32) + + net = SwinIRNet( + upscale=1, + in_channels=3, + img_size=48, + window_size=8, + img_range=1.0, + depths=[6], + embed_dim=60, + num_heads=[6], + mlp_ratio=2, + upsampler='', + resi_connection='1conv') + img = torch.rand(1, 3, 16, 16) + output = net(img) + assert isinstance(output, torch.Tensor) + assert output.shape == (1, 3, 16, 16) + + # x3 model classical SR, initialization and forward (cpu) + net = SwinIRNet( + upscale=3, + in_channels=3, + img_size=16, + window_size=8, + img_range=1.0, + depths=[2], + embed_dim=8, + num_heads=[2], + mlp_ratio=2, + upsampler='pixelshuffle', + resi_connection='1conv') + img = torch.rand(1, 3, 16, 16) + output = net(img) + assert isinstance(output, torch.Tensor) + assert output.shape == (1, 3, 48, 48) + + # x4 model lightweight SR, initialization and forward (cpu) + net = SwinIRNet( + upscale=4, + in_channels=3, + img_size=16, + window_size=8, + img_range=1.0, + depths=[2], + embed_dim=8, + num_heads=[2], + mlp_ratio=2, + ape=True, + upsampler='nearest+conv', + resi_connection='1conv') + output = net(img) + assert isinstance(output, torch.Tensor) + assert output.shape == (1, 3, 64, 64) + + +@pytest.mark.skipif( + 'win' in platform.system().lower() and 'cu' in torch.__version__, + reason='skip on windows-cuda due to limited RAM.') +def test_swinir_cuda(): + net = SwinIRNet( + upscale=4, + in_channels=3, + img_size=16, + window_size=8, + img_range=1.0, + depths=[2], + embed_dim=8, + num_heads=[2], + mlp_ratio=2, + upsampler='pixelshuffledirect', + resi_connection='1conv') + img = torch.rand(1, 3, 16, 16) + + # x4 model lightweight SR forward (gpu) + if torch.cuda.is_available(): + net = net.cuda() + output = net(img.cuda()) + assert isinstance(output, torch.Tensor) + assert output.shape == (1, 3, 64, 64) diff --git a/tests/test_models/test_editors/test_swinir/test_swinir_rstb.py b/tests/test_models/test_editors/test_swinir/test_swinir_rstb.py new file mode 100644 index 0000000000..b5f9c1db1d --- /dev/null +++ b/tests/test_models/test_editors/test_swinir/test_swinir_rstb.py @@ -0,0 +1,25 @@ +# Copyright (c) OpenMMLab. All rights reserved. +import platform + +import pytest +import torch + +from mmedit.models.editors.swinir.swinir_rstb import RSTB + + +@pytest.mark.skipif( + 'win' in platform.system().lower() and 'cu' in torch.__version__, + reason='skip on windows-cuda due to limited RAM.') +def test_rstb(): + + net = RSTB( + dim=6, input_resolution=(8, 8), depth=6, num_heads=6, window_size=8) + + img = torch.randn(1, 64, 6) + output = net(img, (8, 8)) + assert output.shape == (1, 64, 6) + + if torch.cuda.is_available(): + net = net.cuda() + output = net(img.cuda(), (8, 8)) + assert output.shape == (1, 64, 6) diff --git a/tests/test_models/test_editors/test_swinir/test_swinir_utils.py b/tests/test_models/test_editors/test_swinir/test_swinir_utils.py new file mode 100644 index 0000000000..0971960401 --- /dev/null +++ b/tests/test_models/test_editors/test_swinir/test_swinir_utils.py @@ -0,0 +1,26 @@ +# Copyright (c) OpenMMLab. All rights reserved. +import torch + +from mmedit.models.editors.swinir.swinir_utils import (drop_path, to_2tuple, + window_partition, + window_reverse) + + +def test_drop_path(): + x = torch.randn(1, 3, 8, 8) + x = drop_path(x) + assert x.shape == (1, 3, 8, 8) + + +def test_to_2tuple(): + x = 8 + x = to_2tuple(x) + assert x == (8, 8) + + +def test_window(): + x = torch.randn(1, 8, 8, 3) + x = window_partition(x, 4) + assert x.shape == (4, 4, 4, 3) + x = window_reverse(x, 4, 8, 8) + assert x.shape == (1, 8, 8, 3)