context_encoder_pytorch icon indicating copy to clipboard operation
context_encoder_pytorch copied to clipboard
verified publisher icon BoyuanJiang

Use generated random mask region for training (not the central crop)

Open ericshijian opened this issue 8 years ago • 5 comments

Hi, just want to know if I can use the already generated random mask images for training (instead of the center-cropped images)

It seems that you have hard encoded the center position in the training part, so wondered if the positions are really needed.

ericshijian avatar Aug 02 '17 06:08 ericshijian

@ericshijian I have ever tried to train the network using randomly cropped images but it failed to converge.You can contact with the author of that paper for details.

BoyuanJiang avatar Aug 02 '17 08:08 BoyuanJiang

@BoyuanJiang , have you fulfilled random inpainting for now?

micklexqg avatar May 09 '18 02:05 micklexqg

@micklexqg Sorry I have no plan to continue this project for now.

BoyuanJiang avatar May 09 '18 02:05 BoyuanJiang

what a pity!

micklexqg avatar May 09 '18 03:05 micklexqg

@BoyuanJiang , can you share your unfinished code for random inpainting? I am working on this. maybe after i realized it , i can help you update

micklexqg avatar May 12 '18 12:05 micklexqg