Human Pose Estimation Using Per-Point Body Region Assignment

In recent years, the task of human pose estimation has become increasingly important, due to large scale usage, including VR applications, as well higher-level tasks, such behavior understanding. this paper, we introduce a novel two-stage deep learning approach named Segmentation-Guided Pose Estimation (SGPE). The pipeline is based on two neural networks working in sequential fashion, while both models effectively process unorganized point clouds input. First, segmentation network performs pointwise classification into corresponding body regions. next step, cloud with per-point region assignment, forming fourth input channel, passed regression network. This way, local and global features are preserved, helping model fully maintain structure. Our strategy achieves competitive results all examined benchmark datasets, outperforms state-of-the-art methods.