Multi-Angle Models and Lightweight Unbiased Decoding-Based Algorithm for Human Pose Estimation

被引:0
|
作者
He, Jianghai [1 ]
Zhang, Weitong [1 ]
Shang, Ronghua [1 ]
Feng, Jie [1 ]
Jiao, Licheng [1 ]
机构
[1] Xidian Univ, Sch Artificial Intelligence, Key Lab Intelligent Percept & Image Understanding, Minist Educ, Xian 710071, Shaanxi, Peoples R China
基金
中国国家自然科学基金;
关键词
Multi-stage tasks; pose estimation; top-down algorithms; lightweight decoding; multi-angle models; NETWORK;
D O I
10.1142/S0218001423560141
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
When a top-down method is taken to the task of human pose estimation, the accuracy of joint point localization is often limited by the accuracy of human detection. In addition, conventional algorithms commonly encode the image to generate a heat map before processing, but the systematic error in decoding the heat map back to the original image has an impact on the positioning. Therefore, to address the two problems, we propose an algorithm that uses multiple angle models to generate the human boxes and then performs lightweight decoding to recover the image. The new boxes can better fit humans and the recovery error can be reduced. First, we split the backbone network into three sub-networks, the first sub-network is responsible for generating the original human box, the second sub-network is responsible for generating a coarse pose estimation in the boxes, and the third sub-network is responsible for a high-precision pose estimation. In order to make the human box fit the human body better, with only a small number of interfering pixels inside the box, models of the human boxes with multiple rotation angles are generated. The results from the second sub-network are used to select the best human box. Using this human box as input to the third sub-network can significantly improve the accuracy of the pose estimation. Then to reduce the errors arising from image decoding, we propose a lightweight unbiased decoding strategy that differs from traditional methods by combining multiple possible offsets to select the direction and size of the final offset. On the MPII dataset and the COCO dataset, we compare the proposed algorithm with 11 state-of-the-art algorithms. The experimental results show that the algorithm achieves a large improvement in accuracy for a wide range of image sizes and different metrics.
引用
收藏
页数:32
相关论文
共 50 条
  • [1] Heterogeneous heatmap distillation framework based on unbiased alignment for lightweight human pose estimation
    Du, Congju
    Li, Zhenyu
    Zhao, Huijuan
    He, Shuangjiang
    Yu, Li
    IMAGE AND VISION COMPUTING, 2024, 146
  • [2] Improved lightweight human pose estimation algorithm
    Wang Ming-he
    Xu Wang-ming
    Jiang Hao-kun
    CHINESE JOURNAL OF LIQUID CRYSTALS AND DISPLAYS, 2023, 38 (07) : 955 - 963
  • [3] Lightweight human pose estimation network and angle-based action recognition
    Kang, Guang-Yu
    Lu, Zi-Qian
    Lu, Zhe-Ming
    Journal of Network Intelligence, 2020, 5 (04): : 240 - 250
  • [4] UULPN: An ultra-lightweight network for human pose estimation based on unbiased data processing
    Wang, Wenming
    Zhang, Kaixiang
    Ren, Haopan
    Wei, Dejian
    Gao, Yanyan
    Liu, Juncheng
    NEUROCOMPUTING, 2022, 480 : 220 - 233
  • [5] Human pose estimation based on lightweight basicblock
    Yanping Li
    Ruyi Liu
    Xiangyang Wang
    Rui Wang
    Machine Vision and Applications, 2023, 34
  • [6] Human pose estimation based on lightweight basicblock
    Li, Yanping
    Liu, Ruyi
    Wang, Xiangyang
    Wang, Rui
    MACHINE VISION AND APPLICATIONS, 2023, 34 (01)
  • [7] Lightweight human pose estimation algorithm based on polarized self-attention
    Liu, Shengjie
    He, Ning
    Wang, Cheng
    Yu, Haigang
    Han, Wenjing
    MULTIMEDIA SYSTEMS, 2023, 29 (01) : 197 - 210
  • [8] Lightweight human pose estimation algorithm based on polarized self-attention
    Shengjie Liu
    Ning He
    Cheng Wang
    Haigang Yu
    Wenjing Han
    Multimedia Systems, 2023, 29 : 197 - 210
  • [9] Human Pose Estimation Based on Lightweight Multi-Scale Coordinate Attention
    Li, Xin
    Guo, Yuxin
    Pan, Weiguo
    Liu, Hongzhe
    Xu, Bingxin
    APPLIED SCIENCES-BASEL, 2023, 13 (06):
  • [10] Lightweight Multi-Resolution Network for Human Pose Estimation
    Li, Pengxin
    Wang, Rong
    Zhang, Wenjing
    Liu, Yinuo
    Xu, Chenyue
    CMES-COMPUTER MODELING IN ENGINEERING & SCIENCES, 2024, 138 (03): : 2239 - 2255