Making a Challenge Submission¶
Submitting Your Algorithm Container¶
To submit to a challenge:
- Navigate to the Challenge page
 - Click Submit and select the phase you'd like to submit to
 - Choose from your existing Algorithms (you must be listed as an editor)
 - Or click Manage your algorithms to create a new one
 
💡 Tip: You do not need to create a new algorithm for every update — simply upload a new container to your existing algorithm.

⚠️ Please test your algorithm container locally before uploading.
You can find detailed instructions in the container creation and local testing guide.
⚠️ Submission limits are in place to:
- Prevent test set overfitting
 - Ensure fairness among participants
 - Conserve compute resources
 
Read more on this in this blog post by the MIDOG 2021 organizers.
Submission Tips¶
- Ensure your user account is verified before submitting
 - Configure GPU and memory requirements appropriately
 - Be aware that maximum runtime per submission is determined by the challenge phase (typically between 1 minute and 1 hour)
 - Container size should be under 10GB and only include the dependencies you need. Model weights should be uploaded separately.
  Compress with gzip for upload:
bash docker save your_algorithm | gzip -c > YourAlgorithm.tar.gz - 
Container validation may take up to 24 hours. If your container is not marked as active after that, please contact support.
 - 
To update your algorithm:
- Upload a new Docker container image on the Container Management page of your algorithm
 - Or tag a new release if your algorithm is linked to a GitHub repository
 - ⚠️ Updating your algorithm does not automatically create a new submission — you must submit it manually to the challenge phase again
 
 - 
Test before submitting:
- Use the "Try Out Algorithm" feature to upload test data and verify the outputs
 - Make sure the outputs match what is expected by the challenge phase
 - You can review logs and errors on the Results tab of your algorithm page
 
 - 
No internet access during execution:
- All dependencies and model weights must be bundled in your container
 - You can simulate this locally by running:
  
bash docker run --network=none ... - If your container tries to access the internet, it will fail with errors like 
Temporary failure in name resolution 
 - 
Debugging failed submissions:
- Participants do not have access to submission logs (to avoid leaking test set information)
 - Contact the challenge organizers if you need help identifying why your submission failed
 - Tip: most issues can be caught by testing your container with "Try Out Algorithm" first
 
 - 
Be patient — evaluation may take time depending on system load
- If your submission has not been evaluated after 24 hours, feel free to reach out to support