End-to-End Algorithm for Recovering Human 3D Model from Monocular Images

Human 3D pose estimation is a hot-topic in recent years. However, most existing 3D pose estimation algorithms can only recover the human 3D pose skeletons from monocular images. Due to the lack of important information, such as human shape and texture, the results of these algorithms cannot effectively support high-level computer vision tasks. In this paper, we propose an algorithm to recover human 3D model instead of simple human 3D skeleton from monocular images. The proposed algorithm is robust and can provide a series of important information for high-level computer vision tasks, including shape, texture, etc.