Posture estimation and human body analysis combined learning system based on parameter hard sharing

The invention belongs to the technical field of computer vision, and particularly relates to a posture estimation and human body analysis combined learning system based on parameter hard sharing. The system is divided into four parts, namely a multi-scale feature extraction module, a joint learning module, a human body analysis branch and a posture estimation branch. An input image firstly passes through a multi-scale feature extraction module to obtain image features under different scales, and cross-scale feature fusion is carried out; the features of all the scales are sent into a combined learning module, and combined features of posture estimation and human body analysis are obtained; and finally, the combined features of the posture estimation and the human body analysis are sent into a posture estimation branch and a human body analysis branch to obtain a posture estimation and human body analysis result. The invention provides a multi-task joint learning system for posture estimation and human body an