# NVIDIA/flownet2-pytorch

Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks

Repository: https://github.com/NVIDIA/flownet2-pytorch
Canonical: https://ross.abutalabs.com/products/flownet2-pytorch
Language: Python
License: NOASSERTION
License Family: other
Last push: 2026-03-30T19:34:57+00:00

## Health v2 (maintenance only)
Score: 66/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 74, release rhythm 35, longevity 100
- inputs: {"age_days": 3208, "days_push": 156, "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 3289, forks 752 (observed 2026-08-28T04:07:53.866995+00:00)

## What it is
A PyTorch implementation of FlowNet 2.0 for deep-learning-based optical flow estimation, released by NVIDIA. It provides multiple network architectures, custom CUDA layers, dataloaders, pretrained models, and multi-GPU training/inference support.

## Use cases
- estimate optical flow between video frames with deep networks
- train FlowNet2 on custom optical flow datasets
- run optical flow inference with pretrained FlowNet2 weights
- benchmark optical flow models on MPI-Sintel
- run half-precision fp16 optical flow inference on GPU

## When to choose
- you need a proven FlowNet 2.0 implementation with pretrained weights
- you want multi-GPU training and fp16 inference for optical flow
- you are reproducing research results on MPI-Sintel or similar benchmarks

## When to avoid
- you need a modern, actively maintained optical flow model (e.g., RAFT)
- you require half-precision custom CUDA layers, which are unsupported
- you cannot pin to old PyTorch versions (0.4.x)

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, computer-vision, image-processing
- domain: computer-vision, deep-learning, machine-learning
- platform: python
- tags: optical-flow, pytorch, flownet2, cuda-kernels, pretrained-models, video-motion-estimation, gpu, linux

## Member repositories
- NVIDIA/flownet2-pytorch (main) score 66

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:53.866995+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 2026-08-30T07:22:50.277258+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA/flownet2-pytorch (fetched 2026-08-28T04:07:53.866995+00:00, sha a426fd656a4a)
- Data as of 2026-08-30T08:39:29.467469+00:00.