2D human body posture estimation method fusing integrated attention

Abstract

The invention discloses a 2D human body posture estimation method fusing integrated attention, and the method comprises the steps: adding an integrated attention module on the basis of an existing HRNet to construct an HRNet network fusing the integrated attention module, and enabling the HRNet network fusing the integrated attention module to comprise the HRNet and the integrated attention module, the integrated attention module is provided with a first branch, a second branch and a merging branch, the first branch is realized by adopting an average pooling layer, the second branch is realized by adopting an average pooling layer, and the merging branch comprises a concat layer, a first full connection layer, a ReLu activation layer, a second full connection layer and a Sigmoid activation layer; the integrated attention module aggregates information in the channel direction by extracting layer domain feature sets of different receptive field sizes, and the purpose of local feature re-learning is achieved; the method has the advantages of high human body posture estimation speed and high accuracy.

Description

technical field [0001] The invention relates to a 2D human body pose estimation method, in particular to a 2D human body pose estimation method with fusion and integrated attention. Background technique [0002] 2D human pose estimation (Human Pose Estimation, HPE) is a very basic task in the field of computer vision. The auxiliary and preparatory work of the task has been widely used in the fields of intelligent video surveillance, human-computer interaction, automatic driving and intelligent medical treatment. However, due to the small joints of the human body, the poses are changeable, and are usually affected by complex backgrounds and differences in appearance features, such as clothing, body shape, self-occlusion and occlusion caused by human movements, it is difficult to accurately estimate the joints of the human body. position becomes a challenging task. [0003] In recent years, deep learning methods have achieved great success in the field of image recognition, ...

AI Technical Summary

Problems solved by technology

[0004]However, the above two 2D human body pose estimation methods based on deep convolutional neural network models have the following problems: 1. There are parameters in the two deep convolutional neural network models The problem of large amount, slow calculation speed and slow reasoning speed leads to slow estimation of human body pose; 2. Human body joint points account for a small proportion in the image, which is a small target and needs to learn local information of the image, but the two depths Convolutional neural network models do not have the ability to learn advanced local information, which leads to low accuracy of human pose estimation

Images