Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching
FreeMatching improves identity-preserving dense correspondence for image edits that break physical continuity.
FreeMatching is a generalizable dense correspondence framework for image editing and reference-guided generation, where identity can persist despite broken motion or geometry. It combines generative and semantic foundation representations with supervision from classical datasets, tracked videos, and synthetic scenes, plus teacher-guided iterative refinement. A single model substantially improves correspondence on challenging edited image pairs while remaining competitive on classical benchmarks. The authors also use it as a quantitative identity-preservation metric whose scores correlate with human judgment.
- FreeMatching targets identity-preserving matches when physical continuity is broken.
- Heterogeneous supervision covers classical data, tracked video, and synthetic scenes.
- One model improves image-editing correspondence and stays competitive classically.
- Its scores correlate with human judgments of identity preservation.
From the original article
1 min read at arxiv.orgDense correspondence matching has historically been bounded by simplifying spatio-temporal priors, such as smooth motion and rigid geometry. While effective for classical tasks, these assumptions break down in image editing and reference-guided generation (IEG), where transformations can preserve visual identity while breaking physical continuity.
By Luping Liu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.