Single upper limb pose estimation method based on improved stacked hourglass network

Gang Peng,Yuezhi Zheng,Jianfeng Li,Jin Yang,Zhonghua Deng

DOI: https://doi.org/10.48550/arXiv.2004.07456

2020-04-16

Abstract:At present, most high-accuracy single-person pose estimation methods have high computational complexity and insufficient real-time performance due to the complex structure of the network model. However, a single-person pose estimation method with high real-time performance also needs to improve its accuracy due to the simple structure of the network model. It is currently difficult to achieve both high accuracy and real-time performance in single-person pose estimation. For use in human-machine cooperative operations, this paper proposes a single-person upper limb pose estimation method based on an end-to-end approach for accurate and real-time limb pose estimation. Using the stacked hourglass network model, a single-person upper limb skeleton key point detection model was <a class="link-external link-http" href="http://designed.Deconvolution" rel="external noopener nofollow">this http URL</a> was employed to replace the up-sampling operation of the hourglass module in the original model, solving the problem of rough feature maps. Integral regression was used to calculate the position coordinates of key points of the skeleton, reducing quantization errors and calculations. Experiments showed that the developed single-person upper limb skeleton key point detection model achieves high accuracy and that the pose estimation method based on the end-to-end approach provides high accuracy and real-time performance.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to achieve high - precision and real - time estimation of single - person upper - limb postures in human - machine collaborative operations. At present, most high - precision single - person posture estimation methods are difficult to meet the real - time performance requirements due to the complex network model structure and high computational complexity; while those methods with high real - time performance have insufficient precision because of the simple network model structure. Therefore, it is currently very difficult to achieve both high - precision and real - time performance in single - person posture estimation. In response to this challenge, this paper proposes a single - person upper - limb posture estimation method based on an improved stacked hourglass network, aiming to achieve high - precision and real - time upper - limb posture estimation through an end - to - end method.

Specifically, the paper solves the above problems by designing a single - person upper - limb skeleton key - point detection model. This model uses a stacked hourglass network and improves the up - sampling operation in the hourglass module, replacing the original nearest - neighbor interpolation method with deconvolution, which solves the problem of rough feature maps. In addition, the paper also uses integral regression instead of the maximum probability method to calculate the position coordinates of skeleton key - points, reducing quantization errors and computational complexity. Experimental results show that the proposed single - person upper - limb skeleton key - point detection model not only improves the detection precision but also maintains good real - time performance.