This contains code for example submissions to the CLIC challenge. It is based on the VVC baselines provided for the challenge.
The image/
directory contains code that decodes into .png
files as needed for the
image compression tracks. The video/
directory has code decoding into raw .yuv
files,
as needed for the video compression tracks. Note that VVC operates only on .yuv
files,
so the image submission also contains a script that converts to .png
after decoding.
The main entry point in both cases is the decode
script. Both scripts expect the
compressed data (bit streams) in a zip file named bs.zip
(not provided here).
The following Docker images are provided for running the submitted scripts. They are based on PyTorch:
800102672637.dkr.ecr.us-east-1.amazonaws.com/compression (CPU)
800102672637.dkr.ecr.us-east-1.amazonaws.com/compression:gpu (GPU)
The corresponding Dockerfile
s can be found in the docker/
directory.
If you need significant changes to the environment your decoder runs in, please email the CLIC mailing list and request a custom image. When doing so, please mention your team name and provide a Dockerfile.