# zju3dv/LoFTR

Code for "LoFTR: Detector-Free Local Feature Matching with Transformers", CVPR 2021, T-PAMI 2022

Repository: https://github.com/zju3dv/LoFTR
Canonical: https://ross.abutalabs.com/products/loftr
Homepage: https://zju3dv.github.io/loftr/
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Topics: feature-matching, pose-estimation, 3d-vision
Last push: 2024-05-31T06:54:41+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2010, "days_push": 824, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2950, forks 421 (observed 2026-08-28T04:07:31.960587+00:00)

## What it is
LoFTR is a detector-free local image feature matching method using Transformers, released with PyTorch inference and training code plus pretrained indoor/outdoor models. It produces semi-dense pixel-level matches between image pairs, excelling in low-texture, blurred, or repetitive regions.

## Use cases
- match features between two images with transformers
- estimate camera pose from image pairs
- find correspondences in low-texture images
- visual localization for slam or sfm pipelines
- run pretrained image matching model in python
- use loftr via kornia

## When to choose
- you need robust semi-dense matches where classic detectors like SIFT fail
- you want a research-grade, well-cited matcher with pretrained models
- you need image matching integrated into kornia or PyTorch workflows

## When to avoid
- you need real-time matching on CPU or very low-latency edge devices
- you only need lightweight sparse keypoints for a production mobile app
- you want a maintained general-purpose vision toolkit rather than a focused matcher

## Facets
- artifact type: library
- maturity: stable
- function: computer-vision, machine-learning, deep-learning
- domain: computer-vision, image-processing, machine-learning, robotics
- platform: python, cross-platform
- tags: feature-matching, pose-estimation, transformers, 3d-vision, visual-localization, image-correspondence, pytorch, gpu

## Member repositories
- zju3dv/LoFTR (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:31.960587+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 2026-08-30T07:32:43.313354+00:00, confidence not recorded.
  - readme: https://github.com/zju3dv/LoFTR (fetched 2026-08-28T04:07:31.960587+00:00, sha dc1731fa0972)
  - homepage: https://zju3dv.github.io/loftr/ (fetched 2026-08-29T09:47:53.577167+00:00, sha f2057b499931)
- Data as of 2026-08-30T08:39:29.467469+00:00.