# facebookresearch/pytorchvideo

A deep learning library for video understanding research.

Repository: https://github.com/facebookresearch/pytorchvideo
Canonical: https://ross.abutalabs.com/products/pytorchvideo
Homepage: https://pytorchvideo.org/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-05-05T20:32:18+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 80, release rhythm 8, longevity 100
- inputs: {"age_days": 2003, "days_push": 120, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3566, forks 425 (observed 2026-08-28T04:08:09.938959+00:00)

## What it is
PyTorchVideo is a deep learning library from Facebook Research focused on video understanding research, built on PyTorch. It provides reusable, modular components including pretrained video models (model zoo), video datasets, data loaders, and video-specific transforms, with support for accelerated inference on hardware including mobile devices.

## Use cases
- classify actions in videos with pretrained models
- train a video classification model in PyTorch
- run video action detection with SlowFast
- load Kinetics-400 pretrained video models via Torch Hub
- decode and preprocess video clips for deep learning
- deploy video models for accelerated mobile inference
- research video understanding architectures

## When to choose
- you need state-of-the-art pretrained video models like SlowFast or X3D
- you want PyTorch-ecosystem components for video research
- you need efficient video-specific data loading and transforms
- you want reproducible benchmarks for video understanding

## When to avoid
- you need general-purpose video editing or encoding rather than ML
- you work with images only instead of video
- you need a non-PyTorch framework like TensorFlow or JAX
- you need actively developed features - the project sees infrequent releases

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, video-processing, data-science
- domain: deep-learning, computer-vision, machine-learning
- platform: python, cross-platform
- tags: pytorch, video-understanding, model-zoo, action-recognition, video-classification, pretrained-models, torch-hub, video, research, gpu, mobile

## Member repositories
- facebookresearch/pytorchvideo (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:09.938959+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 2026-08-29T18:34:10.057214+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/pytorchvideo (fetched 2026-08-28T04:08:09.938959+00:00, sha 61fe1f10590d)
  - homepage: https://pytorchvideo.org/ (fetched 2026-08-29T09:27:50.842201+00:00, sha 1b493719fef5)
  - site_page: https://pytorchvideo.org/docs/tutorial_overview (fetched 2026-08-29T09:27:50.851458+00:00, sha cf64eed9bd26)
  - site_page: https://pytorchvideo.org/docs/tutorial_overview.html (fetched 2026-08-29T09:27:50.853627+00:00, sha cf64eed9bd26)
  - site_page: https://pytorchvideo.org/docs/tutorial_torchhub_inference (fetched 2026-08-29T09:27:50.855760+00:00, sha 03c6862ca0c1)
  - registry_pypi: https://pypi.org/pypi/pytorchvideo/json (fetched 2026-08-29T09:27:50.857944+00:00, sha 1a408a570cf3)
- Data as of 2026-08-30T08:39:29.467469+00:00.