Greetings, I would like to ask whether there are any restrictions on model training for the BraTS 2025 Challenge. From the rules, it appears that the organizing team may attempt to reproduce our results using the provided model container. I have also noticed that in previous BraTS-related publications, participants often specify the hardware used and the total training time required. To clarify: Are there any limits on the duration or compute resources used for training? Is it required to include training reproducibility information (e.g., total epochs, GPU type, runtime) in the submission? Thank you for your time and support. Best regards, Luke Hsu

Created by To-Liang Hsu LukeHsu1999
Dear Mehdi, Thank you for your help and clarification! May I also ask about the deadline for team formation? And if me and my collaborators registered individual teams, can we still merge our teams together? Thank you! Best regards, Luke
Dear @LukeHsu1999 During the training phase, there are no restrictions on computational resources. It's generally recommended to include your training protocols in your manuscript for the sake of reproducibility and transparency. Nevertheless, for the BraTS challenge specifically, please note that the third (testing) phase requires the submitted containerized algorithm to be evaluated on an unseen dataset. We will release instructions on how to submit a containerized algorithm later. Please let us know if you have any question, /Mehdi

Query on Model Training Restrictions page is loading…