# princeton-vl/RAFT

Repository: https://github.com/princeton-vl/RAFT
Canonical: https://ross.abutalabs.com/products/princeton-vl-raft
Language: Python
License: BSD-3-Clause
License Family: permissive
Last push: 2025-08-24T23:34:28+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 38, release rhythm 35, longevity 100
- inputs: {"age_days": 2350, "days_push": 374, "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 4091, forks 679 (observed 2026-08-28T04:08:34.615218+00:00)

## What it is
Official PyTorch implementation of RAFT (Recurrent All Pairs Field Transforms for Optical Flow), an ECCV 2020 model for estimating dense optical flow between video frames. It includes pretrained models, demos, evaluation scripts, and training pipelines for standard optical flow benchmarks.

## Use cases
- estimate optical flow between two image frames
- run a pretrained optical flow model on a sequence of frames
- evaluate optical flow models on Sintel or KITTI benchmarks
- train an optical flow network on FlyingChairs and FlyingThings3D
- reduce GPU memory usage for correlation lookups with an efficient CUDA implementation

## When to choose
- you need state-of-the-art dense optical flow estimation in Python
- you want a research-grade, well-cited optical flow model with pretrained weights
- you have a CUDA GPU and want to train or fine-tune optical flow models

## When to avoid
- you need real-time optical flow on CPU or mobile devices
- you want a production-ready inference service rather than research code
- you lack a CUDA-capable GPU

## Facets
- artifact type: library
- maturity: stable
- function: computer-vision, machine-learning, deep-learning
- domain: computer-vision, machine-learning, deep-learning
- platform: python
- tags: optical-flow, pytorch, eccv-2020, research-code, pretrained-models, gpu, linux

## Member repositories
- princeton-vl/RAFT (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:34.615218+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 2026-08-29T18:23:21.205554+00:00, confidence not recorded.
  - readme: https://github.com/princeton-vl/RAFT (fetched 2026-08-28T04:08:34.615218+00:00, sha 64fe19a2eafc)
- Data as of 2026-08-30T08:39:29.467469+00:00.