README.md

metadata

license: cc-by-nc-sa-4.0
language:
  - en

🦄 UniCorrn: Unified Correspondence Transformer Across 2D and 3D
CVPR 2026

Prajnan Goswami^1*, Tianye Ding^1*, Feng Liu², Huaizu Jiang^1
1 Visual Intelligence Lab, Northeastern University, ² Adobe Research
^* Equal Contribution

Overview

UniCorrn is the first correspondence model with shared weights that unifies 2D-2D, 2D-3D, and 3D-3D geometric matching with an end-to-end transformer architecture.

This model space contains the large-scale pretrained weights from Stage 1 and Stage 2 training, as well as a modified version of the CroCoV2 weights adapted to match our model attributes.

Quick Start

Check out our Github Repo.

Citation

If you find this repository useful in your research, please consider giving a star ⭐ and a citation

@inproceddings{goswami2026unicorrn,
  title={UniCorrn: Unified Correspondence Transformer Across 2D and 3D},
  author={Goswami, Prajnan and Ding, Tianye and Liu, Feng and Jiang, Huaizu},
  booktitle={CVPR},
  year={2026}
}