A multi-person posture estimation method based on a convolutional neural network

• Main Authors: LI XIAOGUANG, CHE XIN, YAN LUXIN, LI CHANGFENG, YU TIANMIN, CHEN TING, ZHONG SHENG, YANG WEIDONG, ZHANG SONGWEI, ZOU LAMEI, XIONG ZIHUA
• Format: Patent
• Language: Chinese; English
• Published: 28.06.2019
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides a multi-person posture estimation method based on a convolutional neural network. The method includes: inputting the to-be-processed image into the trained multi-person postureestimation network to obtain two groups of data including human body key point position data and human body key point mapping vector data; decoding the data to obtain positions of human body key points and human body center points in the image, mapping the key points to a clustered two-dimensional space through a mapping vector, and then clustering the mapped key points by using a k-means algorithm to indirectly achieve grouping of the original human body key points, analyzing the grouped key points, and finally achieve multi-person posture estimation. The multi-person posture estimation network provided by the invention is composed of a feature extraction network, a feature channel compression module, a human body key point position branch module and a human body key point mapping vectorbranch module, and end-to-