Introduction

CLIC 2024, the 6th Challenge on Learned Image Compression, aims to advance the field of image and video compression using machine learning and computer vision. It is also an opportunity to evaluate and compare end-to-end trained approaches against classical approaches. We encourage the development of novel encoder/decoder architectures, novel ways to control information flow between the encoder and the decoder, new perceptual losses, and new ways to learn quantized representations.

Updates
  • April 8th, 2024. Pictures of the awards ceremony are available here!
  • November 9th, 2023. The test phase is now open. Important note: we found a problem in the MP4-to-PNG conversion script that sometimes caused duplicate frames. It is fixed in this patch. If you use this code for your video track test phase submission, please make sure to apply this fix before you prepare it.
  • October 25th, 2023. Three updates:
    1. The paper submission deadline and validation server closing date is now extended to Nov. 8, in line with the DCC paper submission deadline.
    2. Prizes are allocated as follows: $2000 for the first entry, $1000 for the second entry, and $500 for the third entry in each of the four tracks. Corporate teams will not be eligible for prizes.
    3. To help alleviate issues with methods running out of time, we have doubled the decode time available for video models to 20 hours from 10 hours.
  • For 2024, the challenge will move from CVPR to DCC, and will be part of the regular conference program.
Location

The winners of the compression challenge will be announced at Data Compression Conference, which is held annually in Snowbird, Utah. This year, the conference will take place between 19 and 22 March 2024.

Participation

The challenge has a validation phase and a test phase. Submissions will be ranked on this website during the validation phase. The actual winners will be determined during the test phase. For detailed instructions, refer to the tasks page.

Every submission must be accompanied by a paper, submitted through the regular DCC submission system. For the challenge submission to be accepted, each paper must be:

  • accepted as a talk or poster in the regular conference review process, or
  • at least be a self-contained description of the method, detailed enough to reproduce the results. This would typically mean at least three pages.

Papers accepted as talks or posters will be published on IEEExplore, either full-length or as a one-page abstract, respectively. Authors can opt out of being considered for a talk/full-length paper. In addition, full-length papers for all accepted challenge submissions will be made available on this website.

CLIC events at DCC (preliminary)

Date and time (MT)
Talk/Activity
Speaker

Thursday, 21 March 2024

15:30–15:50
CLIC 2024 introduction and methodology
Johannes Ballé (Google), Luca Versari (Google), and Ross Cutler (Microsoft)

15:50–16:10
A Neural Enhancement Post-Processor with a Dynamic AV1 Encoder Configuration Strategy for CLIC 2024
Darren Ramsook and Anil Kokaram, Trinity College Dublin, Ireland

16:10–16:30
UCVC: A Unified Contextual Video Compression Framework with Joint P-frame and B-frame Coding
Jiayu Yang, Wei Jiang, Yongqi Zhai, Chunhui Yang, and Ronggang Wang, Peking University, China

16:45–19:00
Poster session (joint DCC & CLIC)

~18:00
Awards ceremony

All times are in Mountain Time (Salt Lake City, UT local).

Sponsored by