It is not clear to me how the evaluation of our model will be carried out. Should we upload a docker so that you can run the models and obtain the segmentation masks? Or what should the model output be so that you can obtain the metrics?
Created by Rafther0112 Dear @Rafther0112,
Please consult the instructions page found here: https://www.synapse.org/Synapse:syn68633106/wiki/633396. This will let you know what files should go in the folder.
After your upload your file, and submit to the task, you can then choose to rename your folder or even delete it. You can rename by clicking the folder and clicking the folder or file tools button with the ellipses on the right side of the screen. Use the renaming to help you differentiate for yourself what folder uploads you have put onto synapse.
Thanks,
The LISA 2025 Challenge Organizers Thank you very much!
Now I am not sure how to upload the information for the validation stage. What I understand at the moment is that we have the validation set that you give us in the “Files” location. From this, we get the predictions, and according to the page, we must upload a zip file called “LISALFSEG_predictions.zip” What should go inside the folder? The prediction masks with the same names as the data? How do I differentiate between task2s and task2b when uploading? Hi @Rafther0112,
For the testing phase, please submit a separate Docker container for each task. Whether you use the same model architecture for both tasks or different ones is entirely up to you — we will evaluate each submission independently.
Best regards,
LISA 2025 Challenge Organizers Thank you. I have an additional question: since we are working on two different tasks (taska, taskb), can we train two different models, one for each task? Or should our model segment all structures at once? Hi @Rafther0112,
Please check the "Submissions – What to Submit" section on our challenge website. It provides detailed instructions for each phase:
For the Testing Phase, you will need to upload a Docker container so that we can run your model and generate the segmentation masks on our side.
The page also specifies the expected outputs and format for the evaluation.
Best regards,
LISA 2025 Challenge Organizers