#102102nd of 105 in ML frameworks and libraries
slowfast
PyTorch codebase for training video classification and detection backbones
- Stars
- 7.4k
- License
- Apache-2.0
- Last commit
- Mar 2026
- Language
- Python
7,294 on 1 Apr 20267,436 on 21 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
PySlowFast is FAIR's PyTorch codebase for video understanding research. It implements SlowFast, I3D, C2D, Non-local, X3D, MViTv1/v2 and Rev-ViT/Rev-MViT backbones, plus self-supervised methods such as MaskFeat and MAE for video. It ships a model zoo of baselines and trained weights, and visualization tools for training, evaluation and inference.
Who it is for: Researchers training and evaluating video recognition models in PyTorch
Strengths
- Covers many video backbones (SlowFast, X3D, MViTv2, Rev-ViT) in one codebase
- Includes self-supervised recipes: MaskFeat, MAE for video, contrastive SSL
- Model zoo with baseline results and downloadable trained weights
- Supports multigrid training and PyTorchVideo models and datasets
Weaknesses
- Research code; no tagged release published
- Datasets must be prepared manually following DATASET.md
- No Docker or compose files provided
- README gives no hardware requirements for training
What it needs
- Needs PyTorch, PyTorchVideo
- Models: SlowFast, Slow, C2D, I3D, Non-local
Also in ML frameworks and libraries
See all 105| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 94 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 92 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 91 out of 100 |