inpainting_gmcnn

Image Inpainting via Generative Multi-column Convolutional Neural Networks

Input

(Image from https://github.com/shepnerd/inpainting_gmcnn/blob/master/tensorflow/imgs/paris-streetview_256x256/001.png)

Shape : (1, 256, 256, 3)

Output

Left to right: input, mask, image generated by the network, ground truth

Results on Paris streetview with rectangle masks

Shape : (1, 256, 256, 3)

Results on CelebA-HQ with rectangle masks

Shape : (1, 512, 512, 3)

Results on CelebA-HQ with random strokes

Shape : (1, 512, 512, 3)

Results on Places2 with random strokes

Shape : (1, 512, 680, 3)

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 inpainting_gmcnn.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 inpainting_gmcnn.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --model option, you can specify model type which is selected from "paris-streetview", "celebahq", "celebahq-512", "places2".
(default is paris-streetview)
And it can specify mask type "rect" or "stroke" after the --mask_type option.
(default is rect)

$ python3 inpainting_gmcnn --model paris-streetview --mask_type rect

For CelebA-HQ with rectangle masks.

$ python3 inpainting_gmcnn.py --model celebahq -i celebahq_256x256_001.png

For CelebA-HQ with random strokes.

$ python3 inpainting_gmcnn.py --model celebahq-512 --mask_type stroke -i celebahq_512x512_016.png

For Places2 with random strokes.

$ python3 inpainting_gmcnn.py --model places2 --mask_type stroke -i places2_001.png

Reference

Image Inpainting via Generative Multi-column Convolutional Neural Networks

Framework

TensorFlow

Model Format

ONNX opset=11

Netron

paris-streetview_256x256_rect.onnx.prototxt

celebahq_256x256_rect.onnx.prototxt

celebahq_512x512_rect.onnx.prototxt

celebahq_512x512_freeform.onnx.prototxt

places2_512x680_freeform.onnx.prototxt