🔥🔥🔥 Codalab server for test is online now!🔥🔥🔥

Guideline for Challenge

  • If you encounter any questions or misunderstandings, please feel free to contact us:

    We also set up a QQ group (772975876) for quick communication.

Participation requirements

  • The provided test data is NOT allowed to be used for training.

  • NO additional training data is allowed to train/pretrain the model.

  • The submission description should clearly state the algorithm framework.

Award For Each Track

1st-Place: Certificate + 5000 RMB

2nd-Place: Certificate + 3000 RMB

3rd-Place: Certificate + 2000 RMB

Award For Paper

Best Paper: Certificate + 3000 RMB


Acc1 usually refers to Top-1 Accuracy.The number of samples correctly predicted refers to the number of samples in which the category with the highest confidence of the model is the same as the real category, set to M; The total number of samples, the number of all samples, is set to N.Then Acc1 is M/N. Mean Indicates the average accuracy of a classification. That is, the predicted value for each category is `M_i` If the number of samples for each category is `N_i`, and the total number of categories is

$$\text{Mean} = \frac{1}{l} \sum_{i=1}^{l} \frac{M_i}{N_i}$$

Results Format

We will provide corresponding data sets, participants through training fine-grained video samples, get the corresponding action labels of each video sequence, finally submit a zip package containing a txt file. The txt file should contain two columns, the first for the video sequence name and the second for the predicted action tag number, separated by a space.

Please use the following txt format, name the txt file answer.txt and do not change the order of the video sequence name:

video_sequence_name action_tag_number

video_sequence_name action_tag_number


video_sequence_name action_tag_number


Training dataset release: Mar 28 '24 00:00 CST

Test dataset release: Mar 28 '24 00:00 CST

Results Submission: Mar 28 '24 00:00 CST to Apr 15 '24 midnight CST