The participating teams are expected to upload a final docker container on our privately hosted CodaBench Instance that can be used to evaluate their method on the test set. They can base their container on the codalab-legacy:gpu310 container, that follow a certain folder structure.

Codabench Link: CodaBench
Codabench Guidebook: Guidebook

Temporary File Drop for submissions whether there a problems with CodaBench: Link

We are providing also a local evaluation script that calculates our required metrics on a folder of ground truth files and a folder of prediciton files. Evaluation Script⚙️: GitHub.