34.9K ·GitHub
What is this?
Detectron2 is Facebook AI Research's next-generation library for object detection, segmentation and other visual recognition tasks, with state-of-the-art algorithms.
Detectron2 is Facebook AI Research's next-generation library for object detection, segmentation and other visual recognition tasks, with state-of-the-art algorithms.Is it an app I run, or a library I build on?A library: you build research projects and applications on top of it rather than launch it as a standalone program.
A library: you build research projects and applications on top of it rather than launch it as a standalone program.How do I get it and try my first model?Install it per the installation instructions, then follow the Getting Started guide or the official Colab Notebook, which covers basic usage.
Install it per the installation instructions, then follow the Getting Started guide or the official Colab Notebook, which covers basic usage.Does it run on PyTorch?Yes - it is PyTorch-based: trained models export to TorchScript, a PyTorch format, with Caffe2 export also supported. Check the installation instructions for environment requirements.
Yes - it is PyTorch-based: trained models export to TorchScript, a PyTorch format, with Caffe2 export also supported. Check the installation instructions for environment requirements.Is this for researchers or for production systems?Both - it backs computer vision research projects as well as production applications inside Facebook.
Both - it backs computer vision research projects as well as production applications inside Facebook.What tasks and model architectures does it cover?Object detection and segmentation are the core, plus panoptic segmentation, Densepose, Cascade R-CNN, rotated bounding boxes, PointRend, DeepLab, ViTDet and MViTv2.
Object detection and segmentation are the core, plus panoptic segmentation, Densepose, Cascade R-CNN, rotated bounding boxes, PointRend, DeepLab, ViTDet and MViTv2.How is it different from the original Detectron?It is the next-generation successor of both Detectron and maskrcnn-benchmark, and the headline improvement is speed: it trains much faster.
It is the next-generation successor of both Detectron and maskrcnn-benchmark, and the headline improvement is speed: it trains much faster.Where can I see it in action?The README links a blog post with more demos, plus an interview about the stories behind detectron2.
The README links a blog post with more demos, plus an interview about the stories behind detectron2.What are similar projects?ultralytics, X-AnyLabeling, deep-learning-for-image-processing, VisionHOPE, dinov3, yolov7, Swin-Transformer, CVPR2026-Papers-with-Code
ultralytics, X-AnyLabeling, deep-learning-for-image-processing, VisionHOPE, dinov3, yolov7, Swin-Transformer, CVPR2026-Papers-with-Code