CVPR 2024 Workshop | AIGC Quality Evaluation Challenge starts!

NTIRE (New Trends in Image Restoration and Enhancement) is one of the high-profile international symposiums in the field of computer vision, covering low-level vision tasks and holding related challenges. It is worth mentioning that the NTIRE 2024 AIGC Quality Evaluation Challenge will be jointly organized by the team of professors Liu Xiaohong, Min Xiongkuo, and Zhai Guangtao of Shanghai Jiao Tong University and Huawei Technologies Co., Ltd. The 9th NTIRE Symposium is expected to be held simultaneously with CVPR 2024 on June 18, 2024, and will bring together many researchers to discuss the latest technology trends.

Contest background:

In recent years, with the rise of generative artificial intelligence technology, Vincentian picture and Vincentian video models have continued to emerge. The core goal of this type of algorithm is to generate visual content based on user needs. Therefore, humans' subjective preferences for AIGC have begun to receive widespread attention and have important research significance. However, existing image/video quality evaluation methods have weak correlation with human subjective perception of quality, and it is difficult to accurately assess the degree of matching between generated content and user needs. Therefore, there is an urgent need to propose new image/video quality evaluation methods in order to predict the quality of AIGC, thereby promoting the development of the field of multi-modal generation.

We organized an AIGC quality evaluation challenge in cooperation with the NTIRE seminar. Aims to find network designs/solutions that can produce high-quality results that are highly correlated with image and video subjective quality scores (MOS).

Introduction to the competition:

This challenge uses a new data set called AIGCQA-30K, containing 20,000 images of track 1 and 10,000 videodata for Track 2. Covers 16 mainstream Vincent pictures and 8 Vincent video models. Specifically, the training set contains 70%, the validation set 10%, and the test set 20%. During the model development phase, training and validation sets will be released. Participants can use the training set for model training and submit the results on the verification set to the CodaLab server for evaluation. During the testing phase, competition organizers will release a test set. Participants can use the trained model to predict the quality of test images and submit the prediction results, which will be quantitatively evaluated and ranked by the organizers.

Contest official website and challenge website:

NTIRE official website: https://www.php.cn/link/96276c4db2d39c1529b9fdbfc9e5b2e6

AIGC Quality Evaluation Challenge Website (Image Track): https://www.php.cn/link/b58144d7e90b5a43edcce1ca9e642882

AIGC Quality Evaluation Challenge Website (Video Track) ): https://www.php.cn/link/6ef8783d73140a616fc3d27d7330d503

Schedule:

2023.02.04 Release training set and verification Set

2023.02.04 Release of online verification platform

2023.03.07 Release of test set, close verification platform

2023.03.14 Deadline for submission of test results

2023.03.15 Fact sheet and code/executable submission deadline

2023.03.17 Release of preliminary test results to participants

2023.03.28 Paper submission deadline for challenge entries

2023.06.18 NTIRE Seminar and Challenges, Results and Award Ceremony (tentative)

Contest Requirements:

This challenge is open to the whole society. Individuals, institutions of higher learning, scientific research institutions, enterprises and other personnel can register to participate. Each team can only submit one image quality evaluation and/or one video quality evaluation algorithm for final ranking.

Contest rewards:

Participating teams are free to choose whether to publish competition-related papers, and this choice will not become a necessary condition for participating in the challenge or winning the award. This challenge will invite top-ranked entrants to submit papers to the NTIRE Symposium for peer review. After acceptance, the paper will be published in the 2024 CVPR Symposium Proceedings.

The highest-ranking entrants and those who contribute novel approaches to the competition will be invited to become co-authors of a challenge report paper that will be presented at the 2024 CVPR symposium. The image and video tracks will be ranked separately and bonuses will be distributed.

The prize money for both tracks is (tentatively) as follows:

First place: US$1,000 or a gift certificate of equivalent value

Second place: US$600 or Gift certificate of equivalent value

Third place: US$400 or gift certificate of equivalent value

Host team:

NTIRE2024 AIGC Quality Evaluation Challenge is hosted by Co-organized by Shanghai Jiao Tong University and Huawei Technologies Co., Ltd. Among them, the Jiaotong University team members are composed of professors Liu Xiaohong, Min Xiongkuo, Zhai Guangtao and their postdoctoral fellow Sun Wei, doctoral students Li Chunyi, Gao Yixuan, Cao Yuqin, Zhang Zicheng, and master students Kou Tengchuan and Wu Jile.

Contact information of the competition organizer:

Xiaohong Liu ,

Xiongkuo Min ,

Wei Sun ,

Guangtao Zhai .

Chunyi Li ,

Tengchuan Kou ,

Contest communication group:

Welcome everyone to actively participate. If the group is full or the QR code has expired, please email Contact us to join the group!

The above is the detailed content of CVPR 2024 Workshop | AIGC Quality Evaluation Challenge starts!. For more information, please follow other related articles on the PHP Chinese website!