# zju3dv/MatchAnything

Code for "MatchAnything: Universal Cross-Modality Image Matching with Large-Scale Pre-Training", TPAMI 2026.

Repository: https://github.com/zju3dv/MatchAnything
Canonical: https://ross.abutalabs.com/products/matchanything
License Family: other
Last push: 2026-08-06T05:12:26+00:00

## Health v2 (maintenance only)
Score: 64/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 35, longevity 42
- inputs: {"age_days": 601, "days_push": 27, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1279, forks 41 (observed 2026-08-28T04:04:13.568800+00:00)

## What it is
MatchAnything is a deep learning model for universal cross-modality image matching, released as research code accompanying a TPAMI 2026 paper. It provides pre-trained weights via HuggingFace for matching keypoints across images from different modalities (e.g., sketches, paintings, different sensors).

## Use cases
- match keypoints between images from different modalities
- find corresponding points across sketch and photo
- image matching for 3d reconstruction
- estimate pose between cross-modality image pairs
- register images from different sensors
- feature matching robust to style and modality changes

## When to choose
- you need image matching that generalizes across modalities like sketches, paintings, or different sensors
- you want pre-trained weights for feature matching without training your own model
- you are doing SfM, SLAM, or image registration where standard matchers fail on modality gaps

## When to avoid
- you need training code - it is not yet released
- you need a lightweight classical matcher like SIFT/ORB for same-modality images
- you need a permissively licensed production dependency - no license is specified

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, image-processing, machine-learning, deep-learning
- domain: computer-vision, image-processing, machine-learning, deep-learning
- platform: python, cross-platform
- tags: image-matching, feature-matching, cross-modality, pre-trained-models, research-code, 3d-reconstruction, pose-estimation, gpu

## Member repositories
- zju3dv/MatchAnything (main) score 64

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:13.568800+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 2026-08-30T05:02:43.230891+00:00, confidence not recorded.
  - readme: https://github.com/zju3dv/MatchAnything (fetched 2026-08-28T04:04:13.568800+00:00, sha 6f7afd676973)
- Data as of 2026-08-30T08:39:29.467469+00:00.