where is the mask dataset?

[micklexqg]

Hi, thanks for your work. I have a question, where is the mask dataset and how to generate masks?

[bobqywei]

@micklexqg sorry for the late reply, didn't see this till now.

Previously, I used my own simple Python script to generate 256x256 masks. However, I have found that the original dataset (http://masc.cs.gmu.edu/wiki/uploads/partialconv/mask.zip) used provides better results.

The masks are 512x512 and inverted iirc, so you'll have to modify it yourself to work with the 256x256 Places2 data.

[micklexqg]

@bobqywei , thanks for reply and the given url. I notice that there are two outputs in your 'test.py' srcipt. one is output, another is output2. output is generated using the following model: checkpoint_dict = torch.load(cwd + args.model, map_location="cpu") and the output2 is generated using the following model: checkpoint_dict2 = torch.load(cwd + "/model_e0_i56358.pth", map_location="cpu") so, what is the 'model_e0_i56358.pth' ? is args.model your result model? and 'model_e0_i56358.pth' is the model you want to compare?

[bobqywei]

@micklexqg Hey, I wouldn't pay too much attention to my test script, I simply used it to test individual samples on two trained models at a time. Yes, the parameters of my trained models were stored in .pth files in a separate directory.

[bobqywei]

My inpaint.py file provides a simple GUI that I used in my demo video, you just need a trained model stored in .pth

[micklexqg]

@bobqywei ，thanks