Semantic–Structural Graph Convolutional Networks for Whole-Body Human Pose Estimation

Existing whole-body human pose estimation methods mostly segment the parts of body’s hands and feet for specific processing, which not only splits overall semantics body, but also increases amount calculation complexity model. To address these drawbacks, we designed a novel semantic–structural graph convolutional network (SSGCN) tasks, leverages structure to analyze keypoints through improves accuracy estimation. Firstly, introduced heat-map-based keypoint embedding, encodes position information feature body. Secondly, propose consisting several sets cascaded structure-based layers data-dependent non-local layers. Specifically, proposed method extracts groups constructs high-level abstract body process semantic keypoints. The experimental results showed that our achieved very promising on challenging COCO dataset.