# Parskatt/RoMa

[CVPR 2024] RoMa: Robust Dense Feature Matching; RoMa is the robust dense feature matcher capable of estimating pixel-dense warps and reliable certainties for almost any image pair.

Repository: https://github.com/Parskatt/RoMa
Canonical: https://ross.abutalabs.com/products/parskatt-roma
Homepage: https://parskatt.github.io/RoMa/
Language: Python
License: MIT
License Family: permissive
Topics: 3d-reconstruction, dense-matching, feature-matching, image-matching
Last push: 2026-01-23T11:08:04+00:00

## Health v2 (maintenance only)
Score: 52/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 63, release rhythm 19, longevity 85
- inputs: {"age_days": 1199, "days_push": 222, "days_rel": 331, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1293, forks 137 (observed 2026-08-28T04:04:16.152580+00:00)

## What it is
RoMa (romatch) is a Python library for robust dense feature matching between image pairs, estimating pixel-dense warps and reliable certainties. It is the CVPR 2024 research code built on DINOv2 features with a transformer match decoder.

## Use cases
- match keypoints between two images
- estimate dense warps between image pairs
- compute fundamental matrix from image correspondences
- 3d reconstruction from image pairs
- find correspondences in wide-baseline image pairs
- estimate camera pose from image matches

## When to choose
- you need robust dense correspondences between images with large baseline or viewpoint changes
- you want state-of-the-art feature matching for SfM, SLAM, or 3D reconstruction pipelines
- you need certainty estimates to filter unreliable matches

## When to avoid
- you need real-time matching on CPU or low-end hardware
- you only need simple sparse keypoint matching like SIFT/ORB
- you work outside Python or lack a GPU

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, image-processing, machine-learning, deep-learning
- domain: computer-vision, image-processing, machine-learning
- platform: python
- tags: feature-matching, dense-matching, image-matching, 3d-reconstruction, cvpr-2024, dinov2, warp-estimation, pose-estimation, linux, gpu

## Member repositories
- Parskatt/RoMa (main) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:16.152580+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:54:08.017926+00:00, confidence not recorded.
  - readme: https://github.com/Parskatt/RoMa (fetched 2026-08-28T04:04:16.152580+00:00, sha 4802258d2383)
  - homepage: https://parskatt.github.io/RoMa/ (fetched 2026-08-29T12:11:00.667300+00:00, sha 826e6158dd74)
- Data as of 2026-08-30T08:39:29.467469+00:00.