Animal Pose Estimation on APT-36K (val)
82.3APViTPose++-H
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ViTPose++-HBackbone=ViT-H, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 82.3 | 97.7 | 90.6 | 85.5 | 98.3 | |
| ViTPose-HBackbone=ViT-H, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 82.1 | 98.4 | 91.6 | 84.9 | 98.6 | |
| ViTPose++-LBackbone=ViT-L, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 80.8 | 97.4 | 88.5 | 83.9 | 97.9 | |
| ViTPose-LBackbone=ViT-L, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 80.6 | 98.4 | 90.4 | 83.3 | 98.5 | |
| ViTPose++-BBackbone=ViT-B, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 75.9 | 95.4 | 83.7 | 79.2 | 96 | |
| ViTPose-BBackbone=ViT-B, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 75.5 | 97.4 | 85 | 78.4 | 97.6 | |
| HRNetBackbone=HRNet-w32, Resolution=256x256, Ground truth bounding boxes=true2022.12 | 74.2 | — | — | — | — | |
| HRFormerBackbone=HRFormer-B, Resolution=256x256, Ground truth bounding boxes=true2022.12 | 74.2 | — | — | — | — | |
| ViTPose++-SBackbone=ViT-S, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 74.2 | 94.9 | 82.3 | 77.6 | 95.6 | |
| HRNetBackbone=HRNet-w48, Resolution=256x256, Ground truth bounding boxes=true2022.12 | 74.1 | — | — | — | — | |
| HRFormerBackbone=HRFormer-S, Resolution=256x256, Ground truth bounding boxes=true2022.12 | 71.3 | — | — | — | — | |
| ViTPose-SBackbone=ViT-S, Resolution=256x192, Ground truth bounding boxes=true2022.12 | 70.2 | 94.8 | 77.1 | 73.2 | 95.4 | |
| SimpleBaselineBackbone=ResNet-101, Resolution=256x256, Ground truth bounding boxes=true2022.12 | 69.6 | — | — | — | — | |
| SimpleBaselineBackbone=ResNet-50, Resolution=256x256, Ground truth bounding boxes=true2022.12 | 69.4 | — | — | — | — |