VCIP 2025 Video Coding Challenge

About the Challenge

The VCIP 2025 Low-Bitrate Video Compression Challenge aims to advance video compression technology for extremely limited bandwidth scenarios, with a particular focus on ultra low-bitrates (50kbps and 200kbps) where existing methods suffer significant degradation in quality and efficiency. While mainstream streaming services operate at higher bitrates, real-world applications such as emergency command, disaster relief, and field rescue communications often face severe bandwidth restrictions, resulting in poor video quality. Traditional compression benchmarks have largely emphasized objective metrics at conventional bitrates, leaving a critical gap in research targeting ultra low-bitrate environments and perceptual quality assessment. To address this gap, the challenge emphasizes both extreme bitrate constraints and subjective visual quality evaluation, encouraging innovative AI-powered and perceptually optimized codecs for low-latency, bandwidth-constrained scenarios. The challenge further encourages the exploration of generative AI and foundational models, reflecting recent advances in video diffusions, text-to-video and video synthesis, to fundamentally improve compression efficiency and perceptual quality at ultra low bitrates.

Tracks & Requirements

Note: All complexity constraints are evaluated based on the average computational cost per pixel for 720P resolution video. For each track, the encoding complexity is constrained to be no more than 1.5× the decoding complexity.

Test Dataset

The official challenge dataset consists of 720P video sequences (coming soon), primarily derived from HEVC-CTC datasets.
[Download link coming soon]

Important Dates

• Jun. 30, 2025: Registration deadline. Please send your team name, member list, and institutional affiliation to luguo2014@sjtu.edu.cn and wangjing215@huawei.com.
• July. 15, 2025: Official release of the validation datasets to all registered teams.
• Sept. 30, 2025: Submission deadline for challenge papers.
• Oct. 1, 2025: Release of the final test dataset.
• Oct. 5, 2025: Final deadline for submitting compressed bitstreams and the corresponding decoded videos.
• Oct. 15, 2025: Leaderboard and winning teams announcement.
• Oct. 15, 2025: Notification of acceptance for VCIP challenge papers.
• Dec. 1-4, 2025: Award session at VCIP 2025. Winning teams will receive official certificates and be invited to present at the conference.

Evaluation Criteria

• Preliminary Evaluation: Submissions will be ranked using objective metrics, including PSNR, VMAF, and LPIPS.
• Final Evaluation: Winners will be selected through a human subjective evaluation task. Raters will compare the compressed outputs to the original videos and select their preferred codec based on perceived quality.
• Computational Complexity: The total average per-pixel computation for encoding and decoding in each track must meet the specified requirements, as measured on 720P videos.
• Bitrate Constraint: The average must remain below 50/200kbps, as applicable for each track.
• Bandwidth-constrained: Bitrate will be evaluated in 1-second sliding windows to ensure real-world bandwidth compliance.
• Low-latency: Only low-latency encoding structures are allowed. All frames must be decodable in sequential order.

Awards & Sponsors

• First Place (per track): $2000 + Certificate
• Second Place (per track): $1000 + Certificate
• Best Innovation Award: Certificate
• Server support provided by SJTU & Huawei

Organizers

• Prof. Guo Lu (SJTU) | luguo2014@sjtu.edu.cn
• Jing Wang (Huawei) | wangjing215@huawei.com
• Team: Yunuo Chen (SJTU), Yibo Shi (Huawei), Yuxin Zhao (Huawei)

Contact & Participation

For inquiries, please contact:
Prof. Guo Lu (luguo2014@sjtu.edu.cn)
Jing Wang (wangjing215@huawei.com)
Yunuo Chen (cyril-chenyn@sjtu.edu.cn)
Yibo Shi (shiyibo@huawei.com)
Yuxin Zhao (zhaoyuxin22@huawei.com)

Q & A

Q1 · How will PSNR be computed in this challenge?

The PSNR metric will be computed in the YUV420 domain. We will use a weighting scheme of 6 : 1 : 1 for the Y, U and V components, respectively (i.e., PSNR_Y : PSNR_U : PSNR_V = 6 : 1 : 1).

Q2 · Are there any constraints on the intra-period (GOP) configuration?

There is no official constraint on the intra-period for the challenge. Participants are free to choose any intra-period configuration, including open-GOP structures.