# facebookresearch/SlowFast

PySlowFast: video understanding codebase from FAIR for reproducing state-of-the-art video models.

Repository: https://github.com/facebookresearch/SlowFast
Canonical: https://ross.abutalabs.com/products/slowfast
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-03-16T11:07:10+00:00

## Health v2 (maintenance only)
Score: 65/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 72, release rhythm 35, longevity 100
- inputs: {"age_days": 2570, "days_push": 170, "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 7410, forks 1295 (observed 2026-08-28T04:09:59.742555+00:00)

## What it is
PySlowFast is a PyTorch-based open-source video understanding codebase from Facebook AI Research (FAIR). It provides implementations of state-of-the-art video classification and detection models such as SlowFast Networks, X3D, and Multiscale Vision Transformers (MViT), designed for efficient training and rapid research experimentation.

## Use cases
- train a video classification model on Kinetics
- reproduce state-of-the-art video recognition results
- implement action recognition in videos
- run video action detection on AVA
- pretrain video models with self-supervised learning
- experiment with video transformer architectures like MViTv2
- train efficient video models like X3D

## When to choose
- you need state-of-the-art video classification or detection backbones in PyTorch
- you are doing video understanding research and want to build on published FAIR models
- you want efficient training of video models with multigrid methods
- you need self-supervised video pretraining implementations (MAE, MaskFeat)

## When to avoid
- you need a production-ready video inference service rather than a research codebase
- you work outside video understanding, e.g., image-only classification
- you need a simple high-level API with minimal configuration
- you don't have GPU resources for training large video models

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, video-processing, computer-vision
- domain: deep-learning, computer-vision, machine-learning
- platform: python, cross-platform
- tags: video-understanding, video-classification, action-recognition, pytorch, research, fair, vision-transformers, self-supervised-learning, video, gpu, linux

## Member repositories
- facebookresearch/SlowFast (main) score 65

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:59.742555+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 2026-08-29T17:37:50.390285+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/SlowFast (fetched 2026-08-28T04:09:59.742555+00:00, sha ac08235d0d9a)
  - registry_pypi: https://pypi.org/pypi/slowfast/json (fetched 2026-08-29T08:33:25.815472+00:00, sha 117afd278ec2)
- Data as of 2026-08-30T08:39:29.467469+00:00.