OpenPose AI Models

OpenPose models detect human body, hand, and face keypoints from images or video. They output skeleton representations that can be used for pose-guided image generation, motion analysis, fitness tracking, and animation.

What OpenPose models do

• Body pose estimation — detect 18+ body keypoints (joints, limbs)
• Hand keypoint detection — track individual finger positions
• Face landmark detection — locate facial feature points
• Multi-person detection — handle multiple people in a single frame

OpenPose skeletons are commonly used as conditioning input for ControlNet models, where the detected pose guides image generation to match a specific body position.

When choosing an OpenPose model, check whether it supports the keypoint format your downstream pipeline expects (COCO, BODY_25, etc.), whether it handles occlusion and partial visibility, and how it performs with multiple overlapping people in the frame.