Keypoint Detection Model

Currently FastDeploy supports the deployment of two key point detection methods.

Task	Description	Model Format	Example	Version
Single person key point detection	Deploy PP-TinyPose series models with input images containing only a single person	Paddle	Please refer to tinypose	Release/2.5
Single/multi person key point detection	Deploy the model tandem task of PicoDet + PP-TinyPose, where the input image first passes through the detection model to get an independent portrait sub-image, and then passes through the PP-TinyPose model to detect key points	Paddle	Please refer to det_keypoint_unite	Release/2.5

Prepare Pre-trained Model

Here we provides the following pre-trained models, you can download and use them directly.

Model	Description	Model Format	Version
PP-TinyPose-128x96	Single person key point detection	Paddle	Release/2.5
PP-TinyPose-256x192	Single person key point detection	Paddle	Release/2.5
PicoDet-S-Lcnet-Pedestrian-192x192 + PP-TinyPose-128x96	Single person key point detection tandem configuration	Paddle	Release/2.5
PicoDet-S-Lcnet-Pedestrian-320x320 + PP-TinyPose-256x192	Multi person key point detection tandem configuration	Paddle	Release/2.5