A multi-person pose estimation method based on mask perceptual depth reinforcement learning

Abstract

The invention discloses a multi-person pose estimation method based on mask perceptual depth reinforcement learning. Firstly, a multi-person pose estimation model is constructed. The multi-person poseestimation model is composed of a detection network for acquiring a detection frame and a mask, a depth reinforcement learning network for improving the positioning accuracy, and a single-person poseestimation network. Then the multi-character pose estimation model is trained by training samples. During the test, the images to be detected are input into the trained multi-person pose estimation model to obtain the pose of all the detection frames of the images to be detected. The method of the invention introduces the mask information into the depth reinforcement learning network and the single-person attitude estimation network, improves the effects of the two stages, and solves the problems of gradient disappearance and gradient explosion by using the residual structure. The method of the invention is more competitive than other advanced multi-character posture estimation methods.

Description

technical field [0001] The invention relates to human body pose estimation technology, in particular to a multi-person pose estimation method based on mask perception depth reinforcement learning. Background technique [0002] With the deployment of a large number of multimedia sensors and the wide application of motion capture technologies such as fashion design, clinical analysis, human-computer interaction, behavior recognition, and sports rehabilitation, human body pose estimation has become a hot spot in the multimedia industry. [0003] Recently, significant progress has been made in single-person pose estimation by employing deep learning-based architectures. However, multi-person pose estimation, that is, judging the poses of multiple people in an image, especially estimating individuals in a crowd, remains a challenging task. The main difficulties of this task are as follows: First, the number of people in the image is unknown, and the people may appear in any posi...

AI Technical Summary

Problems solved by technology

However, multi-person pose estimation, i.e. judging the poses of multiple people in an image, especially estimating individuals in a crowd, remains a challenging task

The main difficulties of this task are as follows: First, the number of people in the image is unknown, and the people may appear in any position or in any proportion of the image

Second, there is some type of interaction between the characters in the image, such as occlusion, communication, touch, etc., which makes the estimation more difficult

Third, as the number of people in an image increases, the computational complexity increases, making it a challenge to design efficient algorithms

However, redundant detection results are not conducive to human pose estimation

Images