yangxy
commented
2 years ago https://github.com/yangxy/GPEN/blob/ffce0cc34c9926823c9b24b2ef1322a7fb5c44cd/demo.py#L112 Please comment this line. It's worth noting that GPEN-inpainting is blind inpainting and is only trained based on the provided mask-generation methods. It may not work well in arbitrary scenarios.
The inpainting demo appears to use a white line-pattern. Could you provide guidance on what changes / procedures are required to inpaint on white sections of pre-existing images? For example, I wish to provide an image with missing data (denoted by white color), and the algorithm will correctly recognize the missing section(s) and fill them in.
Now, a series of white lines are drawn over the image and then filled in, but the lines are always the same.