FAQ

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.

How is MS-SSIM calculated?
First, a single MS-SSIM value is calculated for each PNG image. We then compute a weighted average where each PNG is weighted by the number of its pixels.

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.

Do we need to encode the Y, U, and V channels of the video data separately?
You may encode the data however you want. For example, you may encode the Y, U, and V channels of one frame into a single file. However, your decoder needs to output separate PNGs for each channel.

Does the model count towards the target bit-rate?
For the video track, the model and data size is combined to calculate a total size (model_size + data_size / 0.019) which has to be below a threshold. For the image track, the model size is not considered and only the size of the encoded data matters.

What are the specifications of the server the decoder is run on?
Each submission is provided 2 CPUs, 12GB RAM and P100 GPU with 16GB memory. There is also a time limit for each decoder which is set to 5 hours for the validation set (if the test set turns out to be larger, the time limit will be proportionally larger).