MMDetection is an open-source object detection and instance segmentation toolbox developed by the MMDet team, which won the COCO Challenge 2018. The toolbox includes a rich set of methods, components, and modules, and provides weights for over 200 network models. It supports multiple frameworks, offers high efficiency, and is continuously updated. The paper introduces the features of MMDetection, conducts benchmarking studies, and shares best practices for training object detectors. Key features include modular design, support for various frameworks, GPU acceleration, and state-of-the-art performance. The toolbox is available at <https://github.com/open-mmlab/mmdetection>.MMDetection is an open-source object detection and instance segmentation toolbox developed by the MMDet team, which won the COCO Challenge 2018. The toolbox includes a rich set of methods, components, and modules, and provides weights for over 200 network models. It supports multiple frameworks, offers high efficiency, and is continuously updated. The paper introduces the features of MMDetection, conducts benchmarking studies, and shares best practices for training object detectors. Key features include modular design, support for various frameworks, GPU acceleration, and state-of-the-art performance. The toolbox is available at <https://github.com/open-mmlab/mmdetection>.