| Rank | Model |
PCKh-0.5 |
Extra Training Data |
Paper | Code | Year |
| 1 |
PCT (swin-l, test set) |
94.3 |
|
Human Pose as Compositional Tokens
|
gengzigang/pct |
2023 |
| 2 |
Soft-gated Skip Connections |
94.1 |
✓ |
Toward fast and accurate human pose estimation via soft-gated skip connections
|
BB-Repos/BBpose · salinasJJ/BBpose · benjs/hourglass_networks |
2020 |
| 3 |
Cascade Feature Aggregation |
93.9 |
✓ |
Cascade Feature Aggregation for Human Pose Estimation
|
|
2019 |
| 4 |
PCT (swin-b, test set) |
93.8 |
|
Human Pose as Compositional Tokens
|
gengzigang/pct |
2023 |
| 5 |
TransPose |
93.5 |
✓ |
TransPose: Keypoint Localization via Transformer
|
yangsenius/TransPose |
2020 |
| 6 |
UniHCP (FT) |
93.2 |
✓ |
UniHCP: A Unified Model for Human-Centric Perceptions
|
opengvlab/unihcp |
2023 |
| 7 |
4xRSN-50 |
93.0 |
|
Learning Delicate Local Representations for Multi-Person Pose Estimation
|
open-mmlab/mmpose · chenyilun95/tf-cpn · caiyuanhao1998/RSN
· +1 |
2020 |
| 8 |
UniPose |
92.7 |
|
UniPose: Unified Human Pose Estimation in Single Images and Videos
|
bmartacho/UniPose · yangyucheng000/unipose-mindspore |
2020 |
| 9 |
MSPN |
92.6 |
|
Rethinking on Multi-Stage Networks for Human Pose Estimation
|
open-mmlab/mmpose · chenyilun95/tf-cpn · megvii-detection/MSPN
· +4 |
2019 |
| 10 |
Spatial Context |
92.5 |
|
Human Pose Estimation with Spatial Contextual Information
|
|
2019 |
| 11 |
HRNet-W32 |
92.3 |
|
Deep High-Resolution Representation Learning for Human Pose Estimation
|
open-mmlab/mmdetection · PaddlePaddle/PaddleDetection · open-mmlab/mmpose
· +36 |
2019 |
| 11 |
DLCM |
92.3 |
|
Deeply Learned Compositional Models for Human Pose Estimation
|
|
2018 |
| 13 |
Multi-Scale Structure-Aware Network |
92.1 |
|
Multi-Scale Structure-Aware Network for Human Pose Estimation
|
|
2018 |
| 14 |
Pyramid Residual Modules (PRMs) |
92.0 |
|
Learning Feature Pyramids for Human Pose Estimation
|
bearpaw/PyraNet · wanggrun/Learning-Feature-Pyramids · wanggrun/Learning-Feature-Pyramids-For-COCO |
2017 |
| 15 |
Chen et al. ICCV'17 |
91.9 |
|
Adversarial PoseNet: A Structure-aware Convolutional Network for Human Pose Estimation
|
rohitrango/Adversarial-Pose-Estimation · Mind23-2/MindCode-5 |
2017 |
| 16 |
Chou et al. arXiv'17 |
91.8 |
|
Self Adversarial Training for Human Pose Estimation
|
dongzhuoyao/jessiechouuu-adversarial-pose |
2017 |
| 17 |
Multi-Context Attention |
91.5 |
|
Multi-Context Attention for Human Pose Estimation
|
wbenbihi/hourglasstensorlfow · bearpaw/pose-attention |
2017 |
| 17 |
Residual Hourglass +ASR+AHO |
91.5 |
|
Jointly Optimize Data Augmentation and Network Training: Adversarial Data Augmentation in Human Pose Estimation
|
|
2018 |
| 19 |
DU-Net |
91.2 |
|
Quantized Densely Connected U-Nets for Efficient Landmark Localization
|
zhiqiangdon/CU-Net |
2018 |
| 19 |
Stacked hourglass + Inception-resnet |
91.2 |
|
Knowledge-Guided Deep Fractal Neural Networks for Human Pose Estimation
|
Guanghan/GNet-pose |
2017 |
| 19 |
EfficientPose IV |
91.2 |
|
EfficientPose: Scalable single-person pose estimation
|
daniegr/EfficientPose |
2020 |
| 22 |
FPD |
91.1 |
|
Fast Human Pose Estimation
|
ilovepose/fast-human-pose-estimation.pytorch |
2018 |
| 23 |
Integral Regression |
91.0 |
|
Integral Human Pose Regression
|
JimmySuen/integral-human-pose · strawberryfg/c2f-3dhm-human-caffe |
2017 |
| 24 |
Stacked Hourglass Networks |
90.9 |
|
Stacked Hourglass Networks for Human Pose Estimation
|
open-mmlab/mmpose · bearpaw/pytorch-pose · MandyMo/pytorch_HMR
· +43 |
2016 |
| 25 |
DarkPose |
90.6 |
|
Distribution-Aware Coordinate Representation for Human Pose Estimation
|
PaddlePaddle/PaddleDetection · open-mmlab/mmpose · leoxiaobin/deep-high-resolution-net.pytorch
· +3 |
2019 |
| 26 |
TFPose(ResNet-50) |
90.4 |
✓ |
TFPose: Direct Human Pose Estimation with Transformers
|
|
2021 |
| 27 |
Human Pose Estimation |
89.7 |
|
Human pose estimation via Convolutional Part Heatmap Regression
|
1adrianb/human-pose-estimation |
2016 |
| 27 |
Part heatmap regression (ResNet-152) |
89.7 |
|
Human pose estimation via Convolutional Part Heatmap Regression
|
1adrianb/human-pose-estimation |
2016 |
| 29 |
DSNTr ResNet-50@28px |
89.5 |
|
Numerical Coordinate Regression with Convolutional Neural Networks
|
anibali/dsntnn · mansimane/WormML |
2018 |
| 30 |
CU-Net |
89.4 |
|
CU-Net: Coupled U-Nets
|
zhiqiangdon/CU-Net |
2018 |
| 31 |
OpenPose |
88.8 |
|
EfficientPose: Scalable single-person pose estimation
|
daniegr/EfficientPose |
2020 |
| 32 |
Convolutional Pose Machines |
88.52 |
|
Convolutional Pose Machines
|
CMU-Perceptual-Computing-Lab/openpose · open-mmlab/mmpose · shihenw/convolutional-pose-machines-release
· +47 |
2016 |
| 32 |
ResNet-152 + intermediate supervision |
88.52 |
|
DeeperCut: A Deeper, Stronger, and Faster Multi-Person Pose Estimation Model
|
eldar/pose-tensorflow · eldar/deepcut · yttrilab/b-soid
· +13 |
2016 |
| 34 |
Dite-HRNet-30 |
87.6 |
|
Dite-HRNet: Dynamic Lightweight High-Resolution Network for Human Pose Estimation
|
ziyizhang27/dite-hrnet |
2022 |
| 35 |
Tucker T-Net |
87.5 |
|
T-Net: Parametrizing Fully Convolutional Nets with a Single High-Order Tensor
|
|
2019 |
| 36 |
CHPR |
86.4 |
|
Compositional Human Pose Regression
|
anibali/h36m-fetch |
2017 |
| 37 |
Lifshitz et al. |
85.0 |
|
Human Pose Estimation using Deep Consensus Voting
|
|
2016 |
| 38 |
EfficientPose RT |
84.8 |
|
EfficientPose: Scalable single-person pose estimation
|
daniegr/EfficientPose |
2020 |
| 39 |
Matrix |
82.5 |
|
Matrix and tensor decompositions for training binary neural networks
|
|
2019 |
| 40 |
DeepCut |
82.40 |
|
DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation
|
eldar/deepcut · eldar/deepcut-cnn · gsoykan/deepercut-replication
· +1 |
2015 |
| 40 |
QP2 |
82.4 |
|
Bottom-Up and Top-Down Reasoning with Hierarchical Rectified Gaussians
|
peiyunh/rg-mpii |
2015 |
| 42 |
Tompson et al. |
82.0 |
|
Efficient Object Localization Using Convolutional Networks
|
cmu-enyac/Renofeation · yukitsuji/chainer_function |
2014 |
| 43 |
x |
81.3 |
|
Hierarchical binary CNNs for landmark localization with limited resources
|
1adrianb/binary-networks-pytorch |
2018 |
| 43 |
IEF |
81.3 |
|
Human Pose Estimation with Iterative Error Feedback
|
pulkitag/ief |
2015 |
| 45 |
QP1 |
81.1 |
|
Bottom-Up and Top-Down Reasoning with Hierarchical Rectified Gaussians
|
peiyunh/rg-mpii |
2015 |
| 46 |
Improved Binary Network (HourGlass) |
80.9 |
|
Improved training of binary networks for human pose estimation and image recognition
|
1adrianb/binary-networks-pytorch |
2019 |
| 47 |
PCT (swin-l, test set) |
94.3 |
|
Human Pose as Compositional Tokens
|
gengzigang/pct |
2023 |
| 48 |
Soft-gated Skip Connections |
94.1 |
✓ |
Toward fast and accurate human pose estimation via soft-gated skip connections
|
BB-Repos/BBpose · salinasJJ/BBpose · benjs/hourglass_networks |
2020 |
| 49 |
Cascade Feature Aggregation |
93.9 |
✓ |
Cascade Feature Aggregation for Human Pose Estimation
|
|
2019 |
| 50 |
PCT (swin-b, test set) |
93.8 |
|
Human Pose as Compositional Tokens
|
gengzigang/pct |
2023 |
| 51 |
TransPose |
93.5 |
✓ |
TransPose: Keypoint Localization via Transformer
|
yangsenius/TransPose |
2020 |
| 52 |
UniHCP (FT) |
93.2 |
✓ |
UniHCP: A Unified Model for Human-Centric Perceptions
|
opengvlab/unihcp |
2023 |
| 53 |
4xRSN-50 |
93.0 |
|
Learning Delicate Local Representations for Multi-Person Pose Estimation
|
open-mmlab/mmpose · chenyilun95/tf-cpn · caiyuanhao1998/RSN
· +1 |
2020 |
| 54 |
UniPose |
92.7 |
|
UniPose: Unified Human Pose Estimation in Single Images and Videos
|
bmartacho/UniPose · yangyucheng000/unipose-mindspore |
2020 |
| 55 |
MSPN |
92.6 |
|
Rethinking on Multi-Stage Networks for Human Pose Estimation
|
open-mmlab/mmpose · chenyilun95/tf-cpn · megvii-detection/MSPN
· +4 |
2019 |
| 56 |
Spatial Context |
92.5 |
|
Human Pose Estimation with Spatial Contextual Information
|
|
2019 |
| 57 |
HRNet-W32 |
92.3 |
|
Deep High-Resolution Representation Learning for Human Pose Estimation
|
open-mmlab/mmdetection · PaddlePaddle/PaddleDetection · open-mmlab/mmpose
· +36 |
2019 |
| 57 |
DLCM |
92.3 |
|
Deeply Learned Compositional Models for Human Pose Estimation
|
|
2018 |
| 59 |
Multi-Scale Structure-Aware Network |
92.1 |
|
Multi-Scale Structure-Aware Network for Human Pose Estimation
|
|
2018 |
| 60 |
Pyramid Residual Modules (PRMs) |
92.0 |
|
Learning Feature Pyramids for Human Pose Estimation
|
bearpaw/PyraNet · wanggrun/Learning-Feature-Pyramids · wanggrun/Learning-Feature-Pyramids-For-COCO |
2017 |
| 61 |
Chen et al. ICCV'17 |
91.9 |
|
Adversarial PoseNet: A Structure-aware Convolutional Network for Human Pose Estimation
|
rohitrango/Adversarial-Pose-Estimation · Mind23-2/MindCode-5 |
2017 |
| 62 |
Chou et al. arXiv'17 |
91.8 |
|
Self Adversarial Training for Human Pose Estimation
|
dongzhuoyao/jessiechouuu-adversarial-pose |
2017 |
| 63 |
Multi-Context Attention |
91.5 |
|
Multi-Context Attention for Human Pose Estimation
|
wbenbihi/hourglasstensorlfow · bearpaw/pose-attention |
2017 |
| 63 |
Residual Hourglass +ASR+AHO |
91.5 |
|
Jointly Optimize Data Augmentation and Network Training: Adversarial Data Augmentation in Human Pose Estimation
|
|
2018 |
| 65 |
DU-Net |
91.2 |
|
Quantized Densely Connected U-Nets for Efficient Landmark Localization
|
zhiqiangdon/CU-Net |
2018 |
| 65 |
Stacked hourglass + Inception-resnet |
91.2 |
|
Knowledge-Guided Deep Fractal Neural Networks for Human Pose Estimation
|
Guanghan/GNet-pose |
2017 |
| 65 |
EfficientPose IV |
91.2 |
|
EfficientPose: Scalable single-person pose estimation
|
daniegr/EfficientPose |
2020 |
| 68 |
FPD |
91.1 |
|
Fast Human Pose Estimation
|
ilovepose/fast-human-pose-estimation.pytorch |
2018 |
| 69 |
Integral Regression |
91.0 |
|
Integral Human Pose Regression
|
JimmySuen/integral-human-pose · strawberryfg/c2f-3dhm-human-caffe |
2017 |
| 70 |
Stacked Hourglass Networks |
90.9 |
|
Stacked Hourglass Networks for Human Pose Estimation
|
open-mmlab/mmpose · bearpaw/pytorch-pose · MandyMo/pytorch_HMR
· +43 |
2016 |
| 71 |
DarkPose |
90.6 |
|
Distribution-Aware Coordinate Representation for Human Pose Estimation
|
PaddlePaddle/PaddleDetection · open-mmlab/mmpose · leoxiaobin/deep-high-resolution-net.pytorch
· +3 |
2019 |
| 72 |
TFPose(ResNet-50) |
90.4 |
✓ |
TFPose: Direct Human Pose Estimation with Transformers
|
|
2021 |
| 73 |
Human Pose Estimation |
89.7 |
|
Human pose estimation via Convolutional Part Heatmap Regression
|
1adrianb/human-pose-estimation |
2016 |
| 73 |
Part heatmap regression (ResNet-152) |
89.7 |
|
Human pose estimation via Convolutional Part Heatmap Regression
|
1adrianb/human-pose-estimation |
2016 |
| 75 |
DSNTr ResNet-50@28px |
89.5 |
|
Numerical Coordinate Regression with Convolutional Neural Networks
|
anibali/dsntnn · mansimane/WormML |
2018 |
| 76 |
CU-Net |
89.4 |
|
CU-Net: Coupled U-Nets
|
zhiqiangdon/CU-Net |
2018 |
| 77 |
OpenPose |
88.8 |
|
EfficientPose: Scalable single-person pose estimation
|
daniegr/EfficientPose |
2020 |
| 78 |
Convolutional Pose Machines |
88.52 |
|
Convolutional Pose Machines
|
CMU-Perceptual-Computing-Lab/openpose · open-mmlab/mmpose · shihenw/convolutional-pose-machines-release
· +47 |
2016 |
| 78 |
ResNet-152 + intermediate supervision |
88.52 |
|
DeeperCut: A Deeper, Stronger, and Faster Multi-Person Pose Estimation Model
|
eldar/pose-tensorflow · eldar/deepcut · yttrilab/b-soid
· +13 |
2016 |
| 80 |
Dite-HRNet-30 |
87.6 |
|
Dite-HRNet: Dynamic Lightweight High-Resolution Network for Human Pose Estimation
|
ziyizhang27/dite-hrnet |
2022 |
| 81 |
Tucker T-Net |
87.5 |
|
T-Net: Parametrizing Fully Convolutional Nets with a Single High-Order Tensor
|
|
2019 |
| 82 |
CHPR |
86.4 |
|
Compositional Human Pose Regression
|
anibali/h36m-fetch |
2017 |
| 83 |
Lifshitz et al. |
85.0 |
|
Human Pose Estimation using Deep Consensus Voting
|
|
2016 |
| 84 |
EfficientPose RT |
84.8 |
|
EfficientPose: Scalable single-person pose estimation
|
daniegr/EfficientPose |
2020 |
| 85 |
Matrix |
82.5 |
|
Matrix and tensor decompositions for training binary neural networks
|
|
2019 |
| 86 |
DeepCut |
82.40 |
|
DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation
|
eldar/deepcut · eldar/deepcut-cnn · gsoykan/deepercut-replication
· +1 |
2015 |
| 86 |
QP2 |
82.4 |
|
Bottom-Up and Top-Down Reasoning with Hierarchical Rectified Gaussians
|
peiyunh/rg-mpii |
2015 |
| 88 |
Tompson et al. |
82.0 |
|
Efficient Object Localization Using Convolutional Networks
|
cmu-enyac/Renofeation · yukitsuji/chainer_function |
2014 |
| 89 |
x |
81.3 |
|
Hierarchical binary CNNs for landmark localization with limited resources
|
1adrianb/binary-networks-pytorch |
2018 |
| 89 |
IEF |
81.3 |
|
Human Pose Estimation with Iterative Error Feedback
|
pulkitag/ief |
2015 |
| 91 |
QP1 |
81.1 |
|
Bottom-Up and Top-Down Reasoning with Hierarchical Rectified Gaussians
|
peiyunh/rg-mpii |
2015 |
| 92 |
Improved Binary Network (HourGlass) |
80.9 |
|
Improved training of binary networks for human pose estimation and image recognition
|
1adrianb/binary-networks-pytorch |
2019 |
| 93 |
PCT (swin-l, test set) |
94.3 |
|
Human Pose as Compositional Tokens
|
gengzigang/pct |
2023 |
| 94 |
Soft-gated Skip Connections |
94.1 |
✓ |
Toward fast and accurate human pose estimation via soft-gated skip connections
|
BB-Repos/BBpose · salinasJJ/BBpose · benjs/hourglass_networks |
2020 |
| 95 |
Cascade Feature Aggregation |
93.9 |
✓ |
Cascade Feature Aggregation for Human Pose Estimation
|
|
2019 |
| 96 |
PCT (swin-b, test set) |
93.8 |
|
Human Pose as Compositional Tokens
|
gengzigang/pct |
2023 |
| 97 |
TransPose |
93.5 |
✓ |
TransPose: Keypoint Localization via Transformer
|
yangsenius/TransPose |
2020 |
| 98 |
UniHCP (FT) |
93.2 |
✓ |
UniHCP: A Unified Model for Human-Centric Perceptions
|
opengvlab/unihcp |
2023 |
| 99 |
4xRSN-50 |
93.0 |
|
Learning Delicate Local Representations for Multi-Person Pose Estimation
|
open-mmlab/mmpose · chenyilun95/tf-cpn · caiyuanhao1998/RSN
· +1 |
2020 |
| 100 |
UniPose |
92.7 |
|
UniPose: Unified Human Pose Estimation in Single Images and Videos
|
bmartacho/UniPose · yangyucheng000/unipose-mindspore |
2020 |