Multi-Class Segmentation of Aortic Branches and Zones in CTA Banner
Join

Submission Guidelines for AortaSeg24 Challenge Algorithm

Disclaimer: Please note that these instructions are supplemental and intended to assist you in submitting a Docker image to the Grand-Challenge for the AortaSeg24 Challenge. For a comprehensive understanding of the submission process, we highly recommend visiting the official documentation on the Grand-Challenge website.

Docker Image Preparation

Begin by preparing the Docker image or container. Detailed guidelines can be found on the Grand-Challenge website's official documentation. However, we have provided specific instructions for preparing the Docker image for upload to the challenge, available on our GitHub page.

Creating a New Algorithm:

Assuming your model is well-trained and performing optimally, and you have packaged it into a Docker image in ".tar.gz" format, proceed with the following steps:

  1. Submission Initiation: - Navigate to the challenge page and click the Submit button. - On the subsequent screen, select Add a New Algorithm. Note: A maximum of three algorithms may be submitted per participant.
  2. Algorithm Configuration: - Provide a title for your algorithm and an optional description. - Choose whether to utilize GPU resources and specify the required memory size. Available resources include Nvidia T4 GPUs with 16GB of memory, 8 CPU cores, and 32GB of CPU memory.

⚠️ Testing Recommendation:

It is crucial to test your algorithm locally before uploading. Instructions for creating and testing your Docker containers can be found in the documentation. To run the inference file of docker image locally, you can follow the guidelines available on our GitHub page.

Updating an Existing Algorithm:

  1. Accessing the Submission Portal: - Click the Submit icon on the challenge page to access the submission portal.
  2. Algorithm Selection: - From the dropdown menu, select your existing algorithm. To add a new version, navigate to the indicated page by clicking on the "on this page" as shown in blue text in the below figure.
  3. Version Management: - Manage container images for your algorithm, such as "Unet_Based_Baseline_Algo_V1".
  4. Image Upload: - Upload the updated container image in ".tar.gz" format. The upload duration will vary based on your internet connection.
  5. Finalization: - After uploading, click Save. The import status will initialize and commence processing, which may take several minutes to hours.

Evaluation Process:

Once submitted, your model will enter the evaluation queue and undergo algorithm execution. The duration of this process depends on your model's complexity and size.

Key Considerations:

  • Ensure your model can process ten images ranging from 100MB to 200MB, as demonstrated in the training dataset.
  • Test your program locally for functionality.
  • Review the complete tutorial and guidelines for Docker image preparation and submission on the Grand-Challenge website.

We wish you the best of luck with your submission! 🍀