Join CVPR 2026 Challenge on Foundation Models for General CT Image Diagnosis!

Show and Tell
1

Hi everyone — we’ve released datasets for the CVPR 2026 Challenge: Foundation Models for General CT Image Diagnosis, and I wanted to share it with the Hugging Face community.

This benchmark evaluates how well 3D CT foundation models generalize across anatomical regions such as the abdomen and chest, under realistic clinical settings including severe class imbalance.

The challenge includes two tracks:

Task 1 – Linear Probing
Evaluate frozen pretrained representations with a lightweight setup.

Task 2 – Embedding Aggregation Optimization
Build custom heads, training schedules, and adaptation strategies on top of publicly available pretrained weights.

We also designed the challenge to be accessible to teams with different compute budgets. For example, teams with limited resources can participate through the Coreset (10%) track, and Task 2 does not require pretraining from scratch. Official baseline contributions from existing CT foundation models such as 3DINO and SPECTRE are available and more are coming soon!

Key dates

  • Validation submissions: until May 10, 2026

  • Test submissions: May 10–15, 2026

  • Paper deadline: June 1, 2026

If you work on medical imaging, foundation models, representation learning, or efficient adaptation, we’d be excited to have you participate!

Resources

We’d love to see your model on the leaderboard and welcome you to join the challenge. Happy coding!

1 Like