Does my model have to reconstruct images in full resolution or can it be cropped?

The decoder has to produce PNG images where each image has the same resolution as the corresponding image in the validation or test set.

How is PSNR calculated?

We compute a single MSE value by averaging across all RGB channels of all pixels of the whole dataset, and from that calculate a PSNR value.

The evaluation server gives "ERROR: Missing image IMG_20170114_210112.png". What am I doing wrong?

The error means that the decoder failed and did not produce all required files. Make sure that the names of the files match the target files' names exactly.

In which directory should the decoder save images?

The decoder can save images in the current working directory . or in any arbitrary subfolder such as images.

How was the P-frame data generated?

We started from the UGC Dataset and applied the following rules:

  • Videos < 720p are discarded
  • Videos with < 100 frames are discarded

All remaining videos are resized to 720p and the YUV 4:2:0 planes are extracted via:

ffmpeg -i ${VIDEO_NAME}.mkv -vf "extractplanes=y,scale=1280:720" ${VIDEO_NAME}/${VIDEO_NAME}_%05d_y.png ffmpeg -i ${VIDEO_NAME}.mkv -vf "extractplanes=u,scale=640:360" ${VIDEO_NAME}/${VIDEO_NAME}_%05d_u.png ffmpeg -i ${VIDEO_NAME}.mkv -vf "extractplanes=v,scale=640:360" ${VIDEO_NAME}/${VIDEO_NAME}_%05d_v.png