FreeMatching draws on representations from FLUX2-4B, a generative foundation model, and DINOv3, a semantic vision model.
Dense matching can track identity beyond physical continuity
FreeMatching combines generative and semantic representations with varied supervision to match image regions across edits, pose changes and conventional vision tasks.
Chinese Tech
Luping Liu · Bingyi Kang · Yifan Wang · Dong Xu
The University of Hong Kong · ByteDance Seed · Zhejiang University
Research Digest··2 min read
Liu et al.
Why this paper
From ByteDance Seed and 2 others
In one line
FreeMatching learns identity-preserving dense correspondence across image editing and generation tasks by combining foundation representations, heterogeneous supervision, and teacher-guided refinement.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (3 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§