hytseng0509 / CrossDomainFewShot

Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation (ICLR 2020 spotlight)
323 stars 62 forks source link

speed up verifying the code of training on multi-domains #5

Open Franklin-Yao opened 4 years ago

Franklin-Yao commented 4 years ago

Hi, I found a few bugs in your code, for example,

  1. m is not used. https://github.com/hytseng0509/CrossDomainFewShot/blob/b03912ce4a8b003240a9cff4c2aee9020ee16db6/methods/LFTNet.py#L14
  2. There is no 'feat_aug' in input args https://github.com/hytseng0509/CrossDomainFewShot/blob/b03912ce4a8b003240a9cff4c2aee9020ee16db6/methods/LFTNet.py#L32 Could you plz spend some efforts to improve the code and make sure the training on multi-domain works? Thanks.
hytseng0509 commented 4 years ago

Thanks for raising those out! We revise the code accordingly.

Franklin-Yao commented 4 years ago

You said 'We are still verifying the code of training with multiple seen domains.' What's the progress so far? Is this version the same as the code you used in your paper?