Submission📦¶

Last edit 15/07/2024 at 15:53 CET

The challenge is split into 3 phases (Sanity check phase, Validation phase, and Testing phase). Once you have decided that your algorithm is ready for submission follow these steps:

• Create your submission docker container: the example submission is available at the LEOPARD-challenge-submission-example GitHub repo. Follow the template and modify the inference code to output your model's predictions. Include the libraries in requirements.txt. Ensure your model weights and all extra resources are included in the Dockerfile. Build and export your docker container. 
• Test your submission docker container LOCALLY: upon building your algorithm, ensure to test it locally. Submissions on the platform happen in the queue and can take time. We highly recommend you perform all the necessary debugging locally. The necessary scripts to test the correctness of your docker container execution are available on the LEOPARD-challenge-submission-example GitHub repo.
• Sanity check phase: after your algorithms' docker container has been built and thoroughly tested locally, - export in and submit the corresponding .tar.gz file to the Sanity check phase on the grand-challenge.org platform. In this phase, 3 images from the public training set are included. The dockerized algorithm is applied to these images and the example C-index is computed. The submissions to this phase are limited to 3 submissions per day.  We highly recommend you upload the container to the Sanity check phase before uploading it to the Validation phase. The evaluation script for the Sanity check phase will be made available on the LEOPARD-challenge-evaluation-method GitHub repo.
• Validation phase: limited to 5 submissions per team for the whole challenge duration.
• Testing phase:  limited to 1 submission for the whole challenge duration.