The ACCV 2022 International Fine-grained Image Analysis Challenge has officially started, and the 220G large-scale data set is open for download simultaneously!
Competition registration and data set download: http://m6z.cn/5IrWQL
Competition introduction
ACCV 2022 Fine-grained Image Analysis Challenge is an international event hosted by Nanjing University of Science and Technology and the University of Wollongong in Australia, with technical support provided by Jishi Platform.
The fine-grained image analysis involved in this competition is a popular research topic in computer vision and pattern recognition. "Samoyed") to conduct research on visual analysis tasks such as positioning, recognition, and retrieval, which has a wide range of application values in real scenarios. However, due to the small inter-class differences and large intra-class differences among subcategories at the fine-grained level, it becomes a more challenging task to distinguish it from traditional image analysis problems.
In this competition, we will set up two tracks including fine-grained image recognition under network supervision and large-scale fine-grained image hash retrieval to evaluate the recognition/retrieval accuracy of the participating algorithms.
Contest questions
Question 1: Fine-grained recognition of network supervision (100g) (800,000 pictures)
We constructed the Web-iNat5000 dataset, which contains more than 800,000 network training images in 5,000 subcategories. It is the largest network-supervised fine-grained dataset ever, and its categories cover multiple meta-categories, such as plants, insects, birds, reptiles, fungi, protozoa, molluscs, animals, etc. For evaluation, a human labeled dataset containing 150k images is used with a uniform distribution for accuracy comparison.
Question 2: Large-scale fine-grained hash retrieval (120g) (500,000 pictures)
We construct the iNatHash-1000 dataset consisting of 1000 subcategories, each containing 500 images. For each subcategory, 400 images are used as training set and validation set, and the remaining 100 images are used to test the model
schedule
At 10:00 on September 8th, the competition will start, the data set will be opened, and the A list will be opened
List B will open at 10:00 on November 5th
10:00, November 8, the deadline for registration
November 08, 23:59, deadline for submission of results
At 10:00 on November 16th, the results of the B list will be announced
Winners will be announced on December 4th
prize settings
The first international competition honor certificate + 15000 maximum value + ACCV 2022 Workshop oral report
The second international competition honor certificate + 12000 maximum value + ACCV 2022 Workshop oral report
The third international competition honor certificate + 10000 maximum value + ACCV 2022 Workshop oral report Report
Certificate of honor for the fourth place in the international competition + 8,000 maximum value
The certificate of honor for the fifth place in the international competition + 6,000 maximum
value
Participants
The competition is open to the whole society. Individuals, institutions of higher learning, scientific research units, enterprises and other personnel can sign up for the competition. Each player can only join 1 team, and the maximum number of team members is 5
Note: 1) Personnel involved in topic writing and data contact in the organization of the competition are prohibited from participating; 2) Employees of the organizer and co-organizers of the competition can participate in the ranking, but not participate in the award evaluation and receive bonuses.
Judges
ways of registration
1. Official registration start time: 10:00 on September 8, 2022, Beijing time
2. Deadline for registration: 10:00, November 8, 2022, Beijing time
3. Registration link: http://m6z.cn/5IrWQL
4. Wechat communication group for the competition: Please add Ji Xiaoxin (ID: cvmart8) to join the group. Please note when adding: Name-Company/School-ACCV2022
5. Competition Q&A mailbox: [email protected]
6. Competition forum communication: https://cvmart.net/community
7. Workshop official website: https://www.accv2022.org/en/