IMPose: Interactive Multi-person Pose Estimation with Dynamic Correction Propagation
Researchers have developed IMPose, a new interactive tool designed to streamline the process of annotating multi-person human pose in videos. This system utilizes a dual-level tracking mechanism that propagates corrections across frames, significantly reducing manual effort. IMPose demonstrates a strong trade-off between accuracy and efficiency, particularly in scenarios requiring minimal user input, and will be open-sourced. AI
IMPACT Streamlines data annotation for AI training, potentially accelerating development in areas requiring human motion analysis.