Missing Definitions for InpaintGenerator_S() and InpaintGenerator_P() in test.py

[AnkitAI07]

While running test.py, I encountered a problem where the definitions for InpaintGenerator_S() and InpaintGenerator_P() were not found. The relevant code snippet is as follows:

if args.type == 'BSCVR_S': model = net.InpaintGenerator_S().to(device) elif args.type == 'BSCVR_P': model = net.InpaintGenerator().to(device)

Questions:

1. Is there a significant difference between InpaintGenerator() and the missing InpaintGenerator_S() and InpaintGenerator_P() in terms of functionality? Could you please elaborate on this?

2. The supplementary materials mentioned in the paper are not available. Could you provide an estimated date for when these materials will be available or a link where they can be accessed?

Thank you for your assistance.

[LIUTIGHE]

Hi thanks for pointing out those issues, I have updated the model file and test scripts. It should work now.

1. Functionally speaking, S and P don't have significant differences. They are just two kinds of basic self-attention implementation to achieve feature completion via treating the tokens in sequence as vanilla Transformer or divide into patches refering to SwinTransformer.
2. You can check the supplementary materials here, and I have update readme as well.

Thanks again for your checking, to be honest I am working on another urgent project now and I didn't make detailed checking. So if you encounter further problems, feel free to leave your issue here and I would be glad to amend the code for you.