MMDetection
An open-source PyTorch toolkit for training and testing object detection and segmentation models, with GPU operations and an Apache 2.0 license.
MMDetection is a Python toolkit for researchers and developers building object detection and image segmentation systems. Part of OpenMMLab, it combines ready-made model architectures with interchangeable components for custom models. It's open source under Apache 2.0.
The toolkit covers object detection, instance segmentation, panoptic segmentation and semi-supervised detection. You can train and test models or use existing models for inference. Its model zoo includes benchmark results alongside models, so you can compare approaches before choosing one for your work.
Model choices include Faster R-CNN, RetinaNet, YOLOX and DETR for detection, plus Mask R-CNN and YOLACT for instance segmentation. Mask2Former supports segmentation tasks, while Soft Teacher provides a semi-supervised detection approach. RTMDet covers object detection and real-time instance segmentation.
Its modular design lets developers combine model components rather than build an entire detection framework from scratch. Available backbones include ResNet, Swin and ConvNeXt, giving researchers room to test different architectures within the same toolkit.
MMDetection uses PyTorch and depends on MMEngine for model training and MMCV for computer vision functionality. Basic bounding-box and mask operations run on GPUs.
