canqin001

Hi, h_i_s and h_j_t mean the deep features of source and target samples respectively and i and j represent the i-th and j-th index of the sample. We directly extract...

Hi Yiru. I have not tested this yet. Did you try all the domain pairs or simply try one pair?

Hi Yiru. I don't see any bugs in your code. For the SDA setting, I guess the model might be overfitting towards the target domain if few labeled samples are...

I have tried this one (https://github.com/mseitzer/pytorch-fid) to compute FID on COCO. The fid score is around 19 which is still higher than the reported results.

That is a good question. I forgot the gap between train and val set. On val set, the FID is 19.11 and the train set is 12.79 (between val-text-generated images...

Yes. I evaluate the full sets. It takes several hours to go.

I just changed `cmd = f'ffmpeg -y -f image2 -loglevel quiet -framerate {save_fps} -i {frame_dir}/%04d.png -vcodec libx264 -crf 17 -pix_fmt yuv420p {local_path}'` into `cmd = f'/usr/bin/ffmpeg -y -f image2 -loglevel...

Hi @zhengzangw ，请问720p支持的最长frames是多少？我在80G GPU上试了16 frames，720x1280，依然会OOM，非常感谢！

> I think augmentations like affine augmentation used in [this repo](https://github.com/Britefury/self-ensemble-visual-domain-adapt) are helpful. What kinds of tricks have you tried? I just rotate the target training images to gain some...

Thank you so much for your explanation! One more question: how to modify the "def __getitem__(self, ):" to make it work?