2014:Learning Human Pose Estimation Features with Convolutional Networks, Jain etc, ICLR 2014
link: https://arxiv.org/pdf/1312.7302.pdf
2014:DeepPose: Human Pose Estimation via Deep Neural Networks(2014) Google 自上而下
link: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6909610
2016:CPM: Convolutional Pose Machines, Wei etc, CVPR 2016
link: https://arxiv.org/pdf/1602.00134.pdf
2016: Hourglass: Stacked Hourglass Networks for Human Pose Estimation, Newell etc, ECCV 2016
link: https://arxiv.org/pdf/1603.06937.pdf
2016: DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation, Pishchulin etc, CVPR 2016
2016: DeeperCut: A Deeper, Stronger, and Faster Multi-Person Pose Estimation Model, Insafutdinov, ECCV 2016
2017: Multi-Context Attention for Human Pose Estimation, Chu etc, CVPR 2017
2017:OpenPose:Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields, CVPR 2017
2017:Associative Embedding: End-to-End Learning for Joint Detection and Grouping, NIPS 2017
2017:RMPE: Regional Multi-Person Pose Estimation(2017)上海交通大学 自上而下
2018:CPN: Cascaded Pyramid Network for Multi-Person Pose Estimation, CVPR 2017
2018:Deeply Learned Compositional Models for Human Pose Estimation, ECCV 2018
2018:MSPN: Rethinking on Multi-Stage Networks for Human Pose Estimation, Arxiv 2018
2019: HRNet: Deep High-Resolution Representation Learning for Human Pose Estimation CVPR 2019
2020:UniPose: Unified Human Pose Estimation in Single Images and Videos CVPR 2020
2020:Delving Into Unbiased Data Processing for Human Pose Estimation CVPR 2020
2020:HigherHRNet: Scale-Aware Representation Learning for Bottom-Up Human Pose Estimation CVPR 2020
2020:Distribution-Aware Coordinate Representation for Human Pose Estimation CVPR 2020
2020:Mixture Dense Regression for Object Detection and Human Pose Estimation CVPR 2020
2021:Deep Dual Consecutive Network for Human Pose Estimation CVPR 2021
2021:Rethinking the Heatmap Regression for Bottom-up Human Pose Estimation CVPR 2021
2021:FCPose: Fully Convolutional Multi-Person Pose Estimation with Dynamic Instance-Aware Convolutions CVPR
2021:ViPNAS: Efficient Video Pose Estimation via Neural Architecture Search CVPR 2021
其他:
Blaze Pose
PoseNet
Media Pipe
PoseEstimation-TFlite
movenet
视频:
1、2018: Simple Baselines for Human Pose Estimation and Tracking, Xiao etc, ECCV 2018
2、2020: Combining Detection and Tracking for Human Pose Estimation in Videos CVPR 2020 Manchen Wang AWS Rekognition
人体姿态估计常用评估指标
Percentage of Correct Parts (PCP)
检测到正确部位的比例。在PCP中认为一个骨骼(limb)被检测到的标准是,
骨骼的两个关节点的位置与真实关节点的位置的距离小于骨骼的长度的一半。
Percent of Detected Joints (PDJ)
检测到的关节点的比例。在PDJ中认为一个关节被检测到的标准是,
关节的位置与真实位置的距离小于躯干对角点的长度(如left shoulder到right hip的距离)的一个比例值。
Percentage of Correct Keypoints (PCK)
关键点被准确检测的比例。计算检测的关键点与其对应的groundtruth间的归一化距离小于设定阈值的比例。
FLIC 中是以躯干直径作为归一化参考。MPII 中是以头部长度作为归一化参考,即 PCKh。
Average Precision (AP)
平均精确度。类似目标检测,将在真实目标和预测值目标之间的匹配程度的阈值度量
由目标检测中的每个目标检测框的交并比(IOU)改为每个目标的关键点相似度(object keypoint similarity, OKS)。
reference: 人体姿态估计综述_seiyaaa的专栏-CSDN博客_人体姿态估计