PaddleDetection
An open-source object detection toolkit built on PaddlePaddle, with pretrained models, multi-object tracking and server or mobile deployment.
PaddleDetection is a computer vision toolkit for developers building image and video analysis systems. It covers training, adapting and deploying models for uses such as inspection and traffic analysis.
- Object detection and segmentation include pretrained models such as RT-DETR, PP-YOLOE, Faster RCNN and Mask RCNN, alongside face and 3D detection.
- Specialized detection includes PP-YOLOE-SOD for small objects and PP-YOLOE-R for rotated bounding boxes.
- Tracking and pose estimation support following multiple objects and detecting human keypoints, with models including ByteTrack, DeepSORT and PP-TinyPose.
- People and vehicle analysis use the PP-Human and PP-Vehicle toolboxes for applications centered on pedestrians or traffic.
- Model customization lets developers combine model components and reduce model size through pruning, quantization or distillation.
PaddleDetection is open source under Apache 2.0 and uses PaddlePaddle. Deployment choices include Python and C++ applications through Paddle Inference, mobile devices through Paddle Lite, and self-hosted services through Paddle Serving. It also supports ONNX export. PP-PicoDet provides a lightweight detection option for mobile hardware.
