facebookresearch/detectron2

Health v2 · maintenance only

Adoption not part of the score

34688 stars · 7940 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Detectron2 is Facebook AI Research's PyTorch-based library for state-of-the-art object detection, instance/panoptic segmentation, and other visual recognition tasks. It provides a modular platform with pretrained model zoos, training/evaluation tooling, and export to TorchScript or Caffe2 for deployment.

Use cases

• detect objects in images with pretrained models
• train a custom instance segmentation model
• run panoptic segmentation on photos
• fine-tune a Faster R-CNN on my own dataset
• export detection models to TorchScript for production
• estimate human body pose with Densepose
• benchmark object detection baselines

When to choose

• you need state-of-the-art detection or segmentation models in PyTorch
• you want a modular research platform to build vision projects on
• you need pretrained baselines and a model zoo
• you need fast training and production export paths

When to avoid

• you need lightweight real-time detection on edge devices
• you work outside PyTorch (e.g., TensorFlow-only)
• you only need simple image classification
• you want a no-code GUI tool rather than a Python library

Facets

library · maturity stable

machine-learning computer-vision image-processing deep-learning computer-vision machine-learning deep-learning image-processing python object-detection instance-segmentation panoptic-segmentation pytorch model-zoo facebook-ai-research torchscript-export pose-estimation linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="facebookresearch/detectron2")