[CVPR 2026] Dishcovery Mission II Challenge - Third MetaFood Workshop

[Bhalaji Nagarajan]

<<Apologies for cross-posting>>

Dear Community,

We are excited to announce the Dishcovery Mission II Challenge, part of the 3rd MetaFood Workshop at CVPR 2026.

The goal of this challenge is to develop a Vision Language Model that can accurately understand food images and match them to the correct textual descriptions. This is a demanding test of fine grained visual recognition and multi-modal alignment in one of the most diverse domains.

Challenge Highlights

• ~400,000 food image–caption pairs
• Realistic multi-modal noise and fine-grained dish ambiguity
• Focus on efficient and scalable VLM architectures
• Global leader board visibility

Whether you're building the next CLIP, scaling SigLIP, or crafting your own multimodal beast --- this is your chance to stress-test and showcase your models.

Top solutions will be invited to present at the MetaFood Workshop, bringing together researchers working on multimodal AI, food computing, and real-world visual understanding.

---------------------------------------------------------------

Challenge Website: https://dishcoveryvlmchallenge.com/

Submission Portal: https://www.kaggle.com/competitions/dishcovery-mission-ii-cvpr-2026

3rd MetaFood Workshop

Held in conjunction with CVPR 2026

June 3-4th, Denver, Colorado, USA

https://sites.google.com/view/cvpr-metafood-2026

----------------------------------------------------------------

Best regards,
Dishcovery Challenge Organizing Committee