Human Pose Analysis

This book stands at the intersection of computer vision, artificial intelligence, and human kinematics, offering a comprehensive exploration of the principles, methodologies, and applications of human pose analysis in video data. It covers two main aspects: human body pose analysis and human head pose analysis. Human body pose analysis involves estimating the position and orientation of major joints and body parts, such as the head, neck, shoulders, elbows, wrists, hips, knees, and ankles, to capture the entire body posture in 2D or 3D space. In contrast, human head pose analysis focuses solely on the head's orientation, typically estimating the angles of rotation around the yaw, pitch, and roll axes to determine the direction in which a person is looking or tilting their head.

The book is divided into three parts, each detailing recent research in different areas of pose analysis. The first chapter provides an overview of human body and head pose analysis, including the fundamental principles of kinematic representation, as well as commonly used datasets and evaluation metrics. The first part, consisting of Chapters 2 and 3, delves into 2D human body pose analysis. The second part, spanning Chapters 4 through 7, covers the latest advancements in 3D human body pose estimation, focusing on inferring 3D positions and orientations of body joints from 2D images or videos. The third part, covering Chapters 8 and 9, presents recent studies on 3D human head pose analysis, encompassing both 3D head pose estimation and prediction. The final chapter concludes by summarizing the techniques discussed and outlining future research directions and applications in human body and head pose analysis.

Introduces novel concepts of human pose analysis and applications that are suitable for a diverse audience Provides state-of-the-art algorithms and implementation details that cover 2D and 3D dimensions Presents illustrative examples of the methods on diverse datasets that provide various challenging human poses

Autorentext

Songlin Du is an associate professor at the School of Automation, Southeast University, Nanjing, China. He received the B.Sc. degree from China University of Geosciences, Wuhan, China, in 2013; the M.Sc. degree from Lanzhou University, Lanzhou, China, in 2015; a Ph.D. degree in physics from Lanzhou University, Lanzhou, China, in 2019; and a Ph.D. degree in engineering from Waseda University, Tokyo, Japan, in 2019. His research interests include computer vision and machine learning.

Takeshi Ikenaga is a professor at the Graduate School of Information, Production and Systems, Waseda University, Kitakyushu, Japan. He received the B.E. and M.E. degrees in electrical engineering and the Ph.D. degree in information & computer science from Waseda University, Tokyo, Japan, in 1988, 1990, and 2002, respectively. He joined the LSI Laboratories, Nippon Telegraph, and Telephone Corporation (NTT), in 1990, where he had been undertaking research on a real-time MPEG2 encoder chipset, and a highly parallel LSI design for image understanding processing. His current interests are image and video processing systems, which cover video compression, video filtering and video recognition.

Inhalt

Chapter 1. Introduction.- Chapter 2. Bidirectionally Learning Heatmaps for 2D Human Pose Estimation.- Chapter 3. Self-Supervised Multi-Person 2D Human Pose Estimation.- Chapter 4. Bidirectional 2D-3D Transformation for 3D Human Pose Estimation.- Chapter 5. Joint Data Augmentation and Representation for 3D Human Pose Estimation.- Chapter 6. Spatial-Temporal Feature Transform for 3D Human Pose Estimation.- Chapter 7. Real-Time 3D Human Pose Estimation from a Single RGB Image.- Chapter 8. Spatio-Temporal Aggregation for 3D Human Head Pose Estimation.- Chapter 9. Spatial-Temporal Pyramid for 3D Human Head Pose Prediction.- Chapter 10. Conclusion and Outlook.