Human body posture estimation method and system

The invention discloses a human body posture estimation method and system, and the method comprises the steps: carrying out the segmentation extraction of an original image needing the human body posture estimation through a segmentation model composed of a Swin Transform and a region proposal network (RPN), and obtaining an image only containing a person; processing the image only containing the person to obtain a patch feature map; performing channel attention processing on the non-overlapping patch feature map with a random position and a fixed size to obtain a feature vector; processing the patch feature map which is not in a fixed size and can cover the whole image by using spatial attention to obtain a feature vector; mixing the feature vectors obtained by the two types of attention to obtain a patch-based mixed attention feature vector; and fusing the feature vectors and classifying to obtain a posture estimation result. According to the invention, more accurate human body posture estimation is realize