GitBucket
|
|
|||
|---|---|---|---|
| .. | |||
| coco-wholebody | 1 year ago | ||
| ubody2d | 1 year ago | ||
| README.md | 1 year ago | ||
README.md
Top-down heatmap-based pose estimation
Top-down methods divide the task into two stages: object detection, followed by single-object pose estimation given object bounding boxes. Instead of estimating keypoint coordinates directly, the pose estimator will produce heatmaps which represent the likelihood of being a keypoint, following the paradigm introduced in Simple Baselines for Human Pose Estimation and Tracking.
Results and Models
COCO-WholeBody Dataset
Results on COCO-WholeBody v1.0 val with detector having human AP of 56.4 on COCO val2017 dataset
| Model | Input Size | Whole AP | Whole AR | Details and Download |
|---|---|---|---|---|
| HRNet-w48+Dark+ | 384x288 | 0.661 | 0.743 | hrnet_dark_coco-wholebody.md |
| HRNet-w32+Dark | 256x192 | 0.582 | 0.671 | hrnet_dark_coco-wholebody.md |
| HRNet-w48 | 256x192 | 0.579 | 0.681 | hrnet_coco-wholebody.md |
| CSPNeXt-m | 256x192 | 0.567 | 0.641 | cspnext_udp_coco-wholebody.md |
| HRNet-w32 | 256x192 | 0.549 | 0.646 | hrnet_ubody-coco-wholebody.md |
| ResNet-152 | 256x192 | 0.548 | 0.661 | resnet_coco-wholebody.md |
| HRNet-w32 | 256x192 | 0.536 | 0.636 | hrnet_coco-wholebody.md |
| ResNet-101 | 256x192 | 0.531 | 0.645 | resnet_coco-wholebody.md |
| S-ViPNAS-Res50+Dark | 256x192 | 0.528 | 0.632 | vipnas_dark_coco-wholebody.md |
| ResNet-50 | 256x192 | 0.521 | 0.633 | resnet_coco-wholebody.md |
| S-ViPNAS-Res50 | 256x192 | 0.495 | 0.607 | vipnas_coco-wholebody.md |
UBody2D Dataset
Result on UBody val set, computed with gt keypoints.
| Model | Input Size | Whole AP | Whole AR | Details and Download |
|---|---|---|---|---|
| HRNet-w32 | 256x192 | 0.690 | 0.729 | hrnet_ubody-coco-wholebody.md |