Multi-person Pose Estimation on OCHuman (val)
92.8APViTPose-G
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ViTPose-GBackbone=ViTAE-G, Resolution=576x432, Ground truth bounding boxes=true2022.04 | 92.8 | 96.9 | — | — | — | 94 | 97.1 | |
| ViTPose-LBackbone=ViT-L, Resolution=256x192, Ground truth bounding boxes=true2022.04 | 90.9 | 95.8 | — | — | — | 92.3 | 96.7 | |
| ViTPose-HBackbone=ViT-H, Resolution=256x192, Ground truth bounding boxes=true2022.04 | 90.9 | 95.8 | — | — | — | 92.3 | 96.6 | |
| ViTPose-BBackbone=ViT-B, Resolution=256x192, Ground truth bounding boxes=true2022.04 | 88 | 94.8 | — | — | — | 89.6 | 95.9 | |
| Ours-HBackbone=Huge, Cropping Strategy=Ground-truth bounding boxes, Pre-training=COCO-trained2025.05 | 86 | — | — | — | — | — | — | |
| ViTPose++-HBackbone=Huge, Cropping Strategy=Ground-truth bounding boxes, Pre-training=COCO-trained2025.05 | 85.7 | — | — | — | — | — | — | |
| Ours-BBackbone=Base, Cropping Strategy=Ground-truth bounding boxes, Pre-training=COCO-trained2025.05 | 82.2 | — | — | — | — | — | — | |
| ViTPose++-BBackbone=Base, Cropping Strategy=Ground-truth bounding boxes, Pre-training=COCO-trained2025.05 | 81.1 | — | — | — | — | — | — | |
| MIPNetBackbone=HRNet-w48, Resolution=384x288, Ground truth bounding boxes=true2022.04 | 74.1 | 89.7 | — | — | — | 81 | — | |
| CCMBackbone=ResNet-502020.02 | 63.2 | 81.9 | 68.1 | 53.7 | 63.2 | — | — | |
| HRNetBackbone=HRNet-w48, Resolution=384x288, Ground truth bounding boxes=true2022.04 | 62.1 | 76.1 | — | — | — | 65.9 | 78.2 | |
| HRNetBackbone=HRNet-w32, Resolution=384x288, Ground truth bounding boxes=true2022.04 | 60.9 | 76 | — | — | — | 65.1 | 78.2 | |
| BaselineBackbone=ResNet-502020.02 | 59.9 | 79.3 | 66.1 | 64.7 | 59.9 | — | — | |
| SimpleBaselineBackbone=ResNet-152, Resolution=384x288, Ground truth bounding boxes=true2022.04 | 58.8 | 72.7 | — | — | — | 63.1 | 75.7 | |
| HRFormerBackbone=HRFormer-S, Resolution=384x288, Ground truth bounding boxes=true2022.04 | 53.1 | 73.1 | — | — | — | 59.6 | 76.9 | |
| HRFormerBackbone=HRFormer-B, Resolution=384x288, Ground truth bounding boxes=true2022.04 | 50.4 | 71.5 | — | — | — | 58.8 | 76.6 | |
| MIPNetBackbone=HRNet-W48-384, Detector=Faster R-CNN, Paradigm=Top-Down2021.01 | 42 | — | — | — | — | — | — | |
| HGGBackbone=Hourglass, Multi-scale test=true, Single-person pose refinement=false2020.07 | 41.8 | — | — | — | — | — | — | |
| HGGParadigm=Bottom-Up, Multi-scale=x2, x1, x0.52021.01 | 41.8 | — | — | — | — | — | — | |
| LOGO-CAPBackbone=HRNet-W48, Approach=Direct2021.09 | 41.2 | — | — | — | — | — | — | |
| SBLBackbone=ResNet152, Multi-scale test=false, Single-person pose refinement=false2020.07 | 41 | — | — | — | — | — | — | |
| SBLBackbone=ResNet-152, Approach=Top-down2021.09 | 41 | — | — | — | — | — | — | |
| AEBackbone=Hourglass, Multi-scale test=true, Single-person pose refinement=false2020.07 | 40 | — | — | — | — | — | — | |
| AEParadigm=Bottom-Up, Multi-scale=x2, x1, x0.52021.01 | 40 | — | — | — | — | — | — | |
| LOGO-CAPBackbone=HRNet-W32, Approach=Direct2021.09 | 39 | — | — | — | — | — | — | |
| RMPEBackbone=Hourglass, Multi-scale test=false, Single-person pose refinement=false2020.07 | 38.8 | — | — | — | — | — | — | |
| RMPEBackbone=Hourglass, Approach=Top-down2021.09 | 38.8 | — | — | — | — | — | — | |
| DEKRBackbone=HRNet-W48, Approach=Grouping2021.09 | 38.8 | — | — | — | — | — | — | |
| DEKRBackbone=HRNet-W32, Approach=Grouping2021.09 | 37.9 | — | — | — | — | — | — | |
| SBLBackbone=ResNet50, Multi-scale test=false, Single-person pose refinement=false2020.07 | 37.8 | — | — | — | — | — | — | |
| HRNetBackbone=HRNet-W48-384, Detector=Faster R-CNN, Paradigm=Top-Down2021.01 | 37.8 | — | — | — | — | — | — | |
| SBLBackbone=ResNet-50, Approach=Top-down2021.09 | 37.8 | — | — | — | — | — | — | |
| HGGBackbone=Hourglass, Multi-scale test=false, Single-person pose refinement=false2020.07 | 35.6 | — | — | — | — | — | — | |
| HGGBackbone=Hourglass, Approach=Grouping2021.09 | 35.6 | — | — | — | — | — | — | |
| MIPNetBackbone=ResNet101, Detector=YOLO-v3, Paradigm=Top-Down2021.01 | 32.8 | — | — | — | — | — | — | |
| AEBackbone=Hourglass, Multi-scale test=false, Single-person pose refinement=false2020.07 | 32.1 | — | — | — | — | — | — | |
| AEBackbone=Hourglass, Approach=Grouping2021.09 | 32.1 | — | — | — | — | — | — | |
| HCQNetBackbone=Swin-B2023.03 | 31 | — | — | — | — | — | — | |
| BaseNet-DPSBackbone=Swin-B2023.03 | 30.2 | — | — | — | — | — | — | |
| Pose2SegBackbone=R-502023.03 | 28.5 | — | — | — | — | — | — |