Pose Detection - 3.5 English

Vitis AI Library User Guide (UG1354)

Document ID
Release Date
3.5 English

The Pose Detection library is used to detect the posture of the human body. This library includes a neural network that can identify 14 key points on the human body. The input is normally a cropped region that was detected by a pedestrian detection neural network such as SSD or RefineDet available in the Vitis AI Model Zoo. The output is a structure containing the coordinates of each point. The following image shows the result of pose detection.

Figure 1. Pose Detection Example

The following table lists the pose detection model supported by the Vitis AI Library.

Table 1. Pose Detection Model
No Model Name Framework
1 sp_net Caffe
Note: If the input image is arbitrary and you do not know the coordinates of the person, perform a cascaded person detection (pose detection pipeline is required). See the test_jpeg_posedetect_with_ssd.cpp file. The input for test_jpeg_posedetect_ssd can be any image with or without a person in it. If there is a person in the image, this cascaded pipeline will first detect the person with SSD, then send the position of the person as the input for posedetect. If the detection network does not localize any person in the image, posedetect does not run. As test_jpeg_posedetect only performs posedetect, the input image must have at least one person. If you input an image without a person for test_jpeg_posedetect, it will throw an error. See the test_jpeg_posedetect.cpp file.