Источник
CVPR / NTIRE
Дата публикации
11.06.2025
Авторы
Николай Сафонов Алексей Брынцев Андрей Москаленко Дмитрий Куликов Дмитрий Ватолин Radu Timofte Haibo Lei Qifan Gao Qing Luo Yaqing Li Jie Song Shaozhe Hao Meisong Zheng Jingyi Xu Chengbin Wu Jiahui Liu Ying Chen Xin Deng Mai Xu Peipei Liang Jie Ma Junjie Jin Yingxue Pang Fangzhou Luo Kai Chen Shijie Zhao Mingyang Wu Renjie Li Yushen Zuo Shengyun Zhong Zhengzhong Tu
Поделиться

NTIRE 2025 challenge on UGC video enhancement: Methods and results

Аннотация

This paper presents an overview of the NTIRE 2025 Challengeon UGC Video Enhancement. The challenge constructeda set of 150 user-generated content videos withoutreference ground truth, which suffer from real-worlddegradations such as noise, blur, faded colors, compressionartifacts, etc. The goal of the participants was to developan algorithm capable of improving the visual qualityof such videos. Given the widespread use of UGC onshort-form video platforms, this task holds substantial practicalimportance. The evaluation was based on subjectivequality assessment in crowdsourcing, obtaining votesfrom over 8000 assessors. The challenge attracted morethan 25 teams submitting solutions, 7 of which passed thefinal phase with source code verification. The outcomesmay provide insights into the state-of-the-art in UGC videoenhancement and highlight emerging trends and effectivestrategies in this evolving research area. All data, includingthe processed videos and subjective comparison votes andscores, is made publicly available—https://github.com/msu-video-group/NTIRE25_UGC_Video_Enhancement.

Присоединяйтесь к AIRI в соцсетях