The PENGWIN 2026 Challenge consists of three independent tasks (Task 1: PENGWIN-Auto, Task 2: PENGWIN-Interact, and Task 3: PENGWIN-Reduction). Participants are allowed to choose one of them or multiple. Each task has a Preliminary Development Phase and a Final Test Phase. Training data is provided to all participants.
The test data will not be opened to the participants. For final submissions, participants must submit their algorithms as a Docker container on the Grand Challenge platform. Additionally, participants are required to provide a description of their proposed algorithm by creating a new topic in the challenge forum.
During the Preliminary Development Phase, participants can have multiple submissions on the validation data. In the Final Test Phase, only the last run of the submitted Docker container is officially counted to compute the challenge results.
Usage of publicly available data including (open) pre-trained nets is allowed. However, participants must specify and discuss any potential differences in results compared to using only the official PENGWIN 2026 dataset. Any external data used must be publicly available.
Participants are required to use their real names and affiliations on their Grand Challenge profiles. Entries from individuals with multiple or duplicate profiles will result in disqualification. Furthermore, participants collaborating together must form a single Team. Only one designated representative per team is permitted to make submissions, and multiple members of the same team are not allowed to submit algorithms independently.
For Task 2 (PENGWIN-Interact), user interaction is allowed during the training phase, but no manual user interaction is allowed during the inference phase on the hidden test set. The algorithms must function fully automatically by processing the input images alongside the standardized pre-simulated interaction points (clicks) provided by the organizers.
The top three performing teams are required to share their source code on a public repository.
Participants are allowed to publish their results separately after the challenge. The publication embargo period will be 12 months after the announcement of the challenge results.
Algorithm Development: Develop your solution. To accelerate your progress, we provide baseline codes as a reference (TBA soon).
Containerization: Package your algorithm into a Docker container (GC instruction). We provide example templates to help you containerize your model for evaluation (TBA soon).
Form a Team: Create or join a team on Grand-Challenge.
Preliminary Phase (Debugging): Submit your container (GC instruction) to the Preliminary Phase to verify successful execution and debug any deployment issues.
Final Submission: Submit your final Docker container to the Final Test Phase for official ranking.
Method Description: Post an algorithm description in the challenge forum to complete your submission.
Presentation: Top-performing teams will be invited to participate in the MICCAI 2026 satellite event in Strasbourg, France, to present their winning methods.
