Research on 2D Human Pose Estimation Based on Deep Learning

Xialing Wu ； Baorong Zhong

《World Scientific Research Journal》 8卷6期 (2022/06) Pp. 47-53

https://doi.org/10.6911/WSRJ.202206_8(6).0007

摘要

Human pose estimation is one of the basic tasks of computer vision, which can be widely used in action recognition, human‐computer interaction, and so on. Human pose estimation is to locate the position information of human pose joint points through input images or videos. The position of the person's pose in the current state can be estimated. The traditional 2D human pose estimation cannot adapt to the complexity of human joints and the transformation of the environment, so it has great limitations. On the other hand, 2D human pose estimation based on deep learning can achieve accurate joint point positions, and the influence factor of environmental changes is small, so it gets rid of the limitations of traditional methods. In this paper, the bottom‐up Higher‐HRNet algorithm is used to evaluate the human body pose, which can estimate the joint points of the human body when it is occluded.

關鍵字

Deep learning ； Human Pose Estimation ； HRNet ； Higher‐HRNet

參考文獻

FISCHLER M A, ELSCHLAGER R A. The representation and matching of pictorial structures[J]. IEEE Transactions on Computers,1973, 22(1): 67-92.

KIEFEL M, GEHLER P V. Human pose estimation with fields of parts[C]// Proceedings of the 2014 European Conference on Computer Vision, LNCS 8693. Cham: Springer, 2014: 331-346.

TOSHEV A, SZEGEDY C. Deeppose: human pose estimation via deep neural networks[C]// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, Piscataway: IEEE,2014: 1653-1660.

SUN K, XIAO B, LIU D, et al. Deep high-resolution representation learning for human pose estimation[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE, 2019: 5693-5703.

HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation,Cheng etc, CVPR 2020.