Explicit Occlusion Reasoning for Multi-person 3D Human Pose Estimation

被引:10
|
作者
Liu, Qihao [1 ]
Zhang, Yi [1 ]
Bai, Song [2 ]
Yuille, Alan [1 ]
机构
[1] Johns Hopkins Univ, Baltimore, MD USA
[2] ByteDance, Singapore, Singapore
来源
关键词
Human pose estimation; 3D from a single image;
D O I
10.1007/978-3-031-20065-6_29
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Occlusion poses a great threat to monocular multi-person 3D human pose estimation due to large variability in terms of the shape, appearance, and position of occluders. While existing methods try to handle occlusion with pose priors/constraints, data augmentation, or implicit reasoning, they still fail to generalize to unseen poses or occlusion cases and may make large mistakes when multiple people are present. Inspired by the remarkable ability of humans to infer occluded joints from visible cues, we develop a method to explicitly model this process that significantly improves bottom-up multi-person human pose estimation with or without occlusions. First, we split the task into two subtasks: visible keypoints detection and occluded keypoints reasoning, and propose a Deeply Supervised Encoder Distillation (DSED) network to solve the second one. To train our model, we propose a Skeleton-guided human Shape Fitting (SSF) approach to generate pseudo occlusion labels on the existing datasets, enabling explicit occlusion reasoning. Experiments show that explicitly learning from occlusions improves human pose estimation. In addition, exploiting feature-level information of visible joints allows us to reason about occluded joints more accurately. Our method outperforms both the state-of-the-art top-down and bottom-up methods on several benchmarks.
引用
收藏
页码:497 / 517
页数:21
相关论文
共 50 条
  • [1] Multi-Person 3D Pose Estimation With Occlusion Reasoning
    Chen, Xipeng
    Zhang, Junzheng
    Wang, Keze
    Wei, Pengxu
    Lin, Liang
    [J]. IEEE TRANSACTIONS ON MULTIMEDIA, 2024, 26 : 878 - 889
  • [2] Dynamic Graph Reasoning for Multi-person 3D Pose Estimation
    Qiu, Zhongwei
    Yang, Qiansheng
    Wang, Jian
    Fu, Dongmei
    [J]. PROCEEDINGS OF THE 30TH ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA, MM 2022, 2022, : 3521 - 3529
  • [3] Mutual Adaptive Reasoning for Monocular 3D Multi-Person Pose Estimation
    Zhang, Juze
    Wang, Jingya
    Shi, Ye
    Gao, Fei
    Xu, Lan
    Yu, Jingyi
    [J]. PROCEEDINGS OF THE 30TH ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA, MM 2022, 2022, : 1788 - 1796
  • [4] Exploring Severe Occlusion: Multi-Person 3D Pose Estimation with Gated Convolution
    Gu, Renshu
    Wang, Gaoang
    Hwang, Jenq-Neng
    [J]. 2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR), 2021, : 8243 - 8250
  • [5] Multi-person 3D Pose Estimation and Tracking in Sports
    Bridgeman, Lewis
    Volino, Marco
    Guillemaut, Jean-Yves
    Hilton, Adrian
    [J]. 2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW 2019), 2019, : 2487 - 2496
  • [6] Multi-Person 3D Human Pose Estimation from Monocular Images
    Dabral, Rishabh
    Gundavarapu, Nitesh B.
    Mitra, Rahul
    Sharma, Abhishek
    Ramakrishnan, Ganesh
    Jain, Arjun
    [J]. 2019 INTERNATIONAL CONFERENCE ON 3D VISION (3DV 2019), 2019, : 405 - 414
  • [7] VoxelTrack: Multi-Person 3D Human Pose Estimation and Tracking in the Wild
    Zhang, Yifu
    Wang, Chunyu
    Wang, Xinggang
    Liu, Wenyu
    Zeng, Wenjun
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2023, 45 (02) : 2613 - 2626
  • [8] Center point to pose: Multiple views 3D human pose estimation for multi-person
    Liu, Huan
    Wu, Jian
    He, Rui
    [J]. PLOS ONE, 2022, 17 (09):
  • [9] AnimePose: Multi-person 3D pose estimation and animation
    Kumarapu, Laxman
    Mukherjee, Prerana
    [J]. PATTERN RECOGNITION LETTERS, 2021, 147 : 16 - 24
  • [10] Multi-person Absolute 3D Human Pose Estimation with Weak Depth Supervision
    Veges, Marton
    Lorincz, Andras
    [J]. ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING, ICANN 2020, PT I, 2020, 12396 : 258 - 270