Detectron is an object detection and instance segmentation research framework that popularized many modern detection models in a single, reproducible codebase. Built on Caffe2 with custom CUDA/C++ operators, it provided reference implementations for models like Faster R-CNN, Mask R-CNN, RetinaNet, and Feature Pyramid Networks. The framework emphasized a clean configuration system, strong baselines, and a “model zoo” so researchers could compare results under consistent settings. It includes training and evaluation pipelines that handle multi-GPU setups, standard datasets, and common augmentations, which helped standardize experimental practice in detection research. Visualization utilities and diagnostic scripts make it straightforward to inspect predictions, proposals, and losses while training. Although the project has since been superseded by Detectron2, the original Detectron remains a historically important, reproducible reference that still informs many productions.
Features
- Reference implementations for Faster R-CNN, Mask R-CNN, RetinaNet, FPN, and keypoint heads
- Model zoo with pretrained weights and standardized evaluation recipes
- YAML-driven config system for experiments and ablations
- Multi-GPU training and efficient Caffe2/CUDA operators
- Dataset loaders and utilities for COCO and other common benchmarks
- Visualization tools for detections, masks, and training diagnostics