PoseNet
lastmod 2026-11-10created 2019-03-013 edits
PoseNet is a deep learning model that estimates human poses in real time in the browser with TensorFlow.js. From an image or webcam video, it returns the positions and confidence scores of 17 keypoints such as the nose, eyes, shoulders, elbows, and knees, and it has a single-pose mode that finds one person and a multi-pose mode that finds multiple people. It is notable for using MobileNet or ResNet as the backbone and running on the client without a server, and MoveNet and BlazePose were later provided as successor models.