NTIRE2024_GoodGame

This repository is the official NTIRE 2024 implementation of Team GoodGame in Stereo Image Super-Resolution Challenge - Track 2 Constrained SR & Realistic Degradation. The restoration results of the testing images can be downloaded from here. Our pretrained models can be downloaded from here.

Usage

Installation environment

python 3.8
pytorch 1.11.0
cuda 11.3

git clone git@github.com:Yuqi-Miao/NTIRE_GoodGame_track2.git
cd NTIRE_GoodGame_track2
pip install -r requirements.txt
python setup.py develop --no_cuda_ext

Modify the configuration file options/train/CVHSSR_Sx4.yml and options/test/CVHSSR_Sx4.yml as follows:

Train
dataroot_gt: ./data/Flickr1024/trainx4 # replace your dataset path
dataroot_lq: ./data/Flickr1024/trainx4 # replace your dataset path

Test
dataroot_gt: ./data/Flickr1024/Stereo_test/KITTI2012/hr # replace your dataset path
dataroot_lq: ./data/Flickr1024/Stereo_test/KITTI2012/lr_x4 # replace your dataset path

Train

python -m torch.distributed.launch --nproc_per_node=2 --master_port=4329 basicsr/train.py -opt options/train/CVHSSR_Sx4.yml --launcher pytorch

infer

Modify the configuration file options/test/CVHSSR_Sx4.yml as follows:

pretrain_network_g: ../pretrain/besk_ckpt.pth # replace your model checkpoint path

Then you can infer

python basicsr/demo_ssr.py -opt options/test/CVHSSR_Sx4.yml --img_pair_paths ntire_test_stereo_image_path --output_dir output_image_path

Parameters

python basicsr/models/archs/CVHSSR_arch.py