Hi, I was looking at the example repository, especially the testing one. For task2, I think we use the best model from each of the clients and test on their own dataset. Now, the dataset is supposed to consist of 3 centers in the training phase and 1 unseen 4th center which would be used for testing. Ref: https://gitlab.com/nct_tso_public/challenges/miccai2024/FedSurg24/-/blob/main/src_team_1/testing/src/main.py?ref_type=heads In the link provided above, I see that for task_2, 4 centers are used. But the 4th center is not used in training and hence wouldn't have the best model. Can you please comment on it? Thank you. Santhi

Created by Santhi Raj Kolamuri SRajKolamuri
Hi, In Task 1, the global model is tested on the 4th center, while in Task 2, each center?s test set is evaluated using its respective model (only centers 1, 2, and 3). This setup should also be reflected in the GitLab repository. However, since the repository was created before I had access to the data, it currently includes 5 centers instead of 4. Best regards, Max

Task 2: 4 centers in the example code page is loading…