#2121st of 22 in ML frameworks and libraries

detectron2

PyTorch library for object detection and segmentation research

Stars
34.9k
License
Apache-2.0
Last commit
Sep 2026
Last release
Nov 2021
Language
Python
Stars, last 29 weeks+831
25 Mar 2026: 34,067 stars1 Apr 2026: 34,093 stars8 Apr 2026: 34,123 stars15 Apr 2026: 34,159 stars22 Apr 2026: 34,200 stars29 Apr 2026: 34,290 stars6 May 2026: 34,317 stars13 May 2026: 34,335 stars20 May 2026: 34,365 stars27 May 2026: 34,395 stars3 Jun 2026: 34,416 stars10 Jun 2026: 34,441 stars17 Jun 2026: 34,464 stars24 Jun 2026: 34,485 stars1 Jul 2026: 34,509 stars8 Jul 2026: 34,524 stars15 Jul 2026: 34,540 stars22 Jul 2026: 34,563 stars29 Jul 2026: 34,584 stars5 Aug 2026: 34,600 stars12 Aug 2026: 34,616 stars19 Aug 2026: 34,634 stars26 Aug 2026: 34,655 stars2 Sept 2026: 34,665 stars9 Sept 2026: 34,687 stars16 Sept 2026: 34,698 stars23 Sept 2026: 34,718 stars30 Sept 2026: 34,736 stars7 Oct 2026: 34,757 stars14 Oct 2026: 34,898 stars

34,067 on 25 Mar 202634,898 on 14 Oct 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Detectron2 is Facebook AI Research's PyTorch library for object detection and segmentation, successor to Detectron and maskrcnn-benchmark. It covers panoptic segmentation, DensePose, Cascade R-CNN, rotated boxes, PointRend, DeepLab, ViTDet and MViTv2, and ships a Model Zoo of baselines. Models can be exported to TorchScript or Caffe2.

Who it is for: Computer vision researchers training detection and segmentation models

Strengths

  • Wide algorithm coverage: panoptic segmentation, DensePose, PointRend, ViTDet, MViTv2, rotated boxes
  • Model Zoo with downloadable baselines and trained weights
  • Export to TorchScript or Caffe2 for deployment
  • Modular design for building research projects on top; Apache-2.0

Weaknesses

  • Last tagged release is 2021-11-15, though commits continue
  • Not a model server; no inference API or UI included
  • Install steps live in external docs, not the README
  • Hardware requirements are not stated in the README

What it needs

  • Docker + Compose
  • Needs PyTorch
  • Models: Mask R-CNN, Cascade R-CNN, Panoptic segmentation, DensePose, PointRend

Also in ML frameworks and libraries

See all 22
Also in ML frameworks and libraries
RankProjectScore
1transformersModel-definition library for text, vision, audio and multimodal models167.7k starsApache-2.094 out of 100
2pytorchPython tensor library with GPU support and tape-based autograd for deep learning104.7k starscustom license93 out of 100
3ultralyticsPython library and CLI for training and running YOLO vision models62.5k starsAGPL-3.088 out of 100
#4supervisionPython toolkit for loading, annotating and counting computer vision detections Live demo ↗ (opens in a new tab)51.2k starsMIT84 out of 100
#5pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts37.2k starsApache-2.082 out of 100