mmpose/configs/face_2d_keypoint/topdown_heatmap/README.md

58 lines
4.5 KiB
Markdown

# 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](http://openaccess.thecvf.com/content_ECCV_2018/html/Bin_Xiao_Simple_Baselines_for_ECCV_2018_paper.html).
<div align=center>
<img src="https://user-images.githubusercontent.com/15977946/146522977-5f355832-e9c1-442f-a34f-9d24fb0aefa8.png" height=400>
</div>
## Results and Models
### 300W Dataset
Results on 300W dataset
| Model | Input Size | NME<sub>*common*</sub> | NME<sub>*challenge*</sub> | NME<sub>*full*</sub> | NME<sub>*test*</sub> | Details and Download |
| :---------: | :--------: | :--------------------: | :-----------------------: | :------------------: | :------------------: | :---------------------------------------: |
| HRNetv2-w18 | 256x256 | 2.92 | 5.64 | 3.45 | 4.10 | [hrnetv2_300w.md](./300w/hrnetv2_300w.md) |
### AFLW Dataset
Results on AFLW dataset
| Model | Input Size | NME<sub>*full*</sub> | NME<sub>*frontal*</sub> | Details and Download |
| :--------------: | :--------: | :------------------: | :---------------------: | :-------------------------------------------------: |
| HRNetv2-w18+Dark | 256x256 | 1.35 | 1.19 | [hrnetv2_dark_aflw.md](./aflw/hrnetv2_dark_aflw.md) |
| HRNetv2-w18 | 256x256 | 1.41 | 1.27 | [hrnetv2_aflw.md](./aflw/hrnetv2_aflw.md) |
### COCO-WholeBody-Face Dataset
Results on COCO-WholeBody-Face val set
| Model | Input Size | NME | Details and Download |
| :--------------: | :--------: | :----: | :----------------------------------------------------------------------------------------------: |
| HRNetv2-w18+Dark | 256x256 | 0.0513 | [hrnetv2_dark_coco_wholebody_face.md](./coco_wholebody_face/hrnetv2_dark_coco_wholebody_face.md) |
| SCNet-50 | 256x256 | 0.0567 | [scnet_coco_wholebody_face.md](./coco_wholebody_face/scnet_coco_wholebody_face.md) |
| HRNetv2-w18 | 256x256 | 0.0569 | [hrnetv2_coco_wholebody_face.md](./coco_wholebody_face/hrnetv2_coco_wholebody_face.md) |
| ResNet-50 | 256x256 | 0.0582 | [resnet_coco_wholebody_face.md](./coco_wholebody_face/resnet_coco_wholebody_face.md) |
| HourglassNet | 256x256 | 0.0587 | [hourglass_coco_wholebody_face.md](./coco_wholebody_face/hourglass_coco_wholebody_face.md) |
| MobileNet-v2 | 256x256 | 0.0611 | [mobilenetv2_coco_wholebody_face.md](./coco_wholebody_face/mobilenetv2_coco_wholebody_face.md) |
### COFW Dataset
Results on COFW dataset
| Model | Input Size | NME | Details and Download |
| :---------: | :--------: | :--: | :---------------------------------------: |
| HRNetv2-w18 | 256x256 | 3.48 | [hrnetv2_cofw.md](./cofw/hrnetv2_cofw.md) |
### WFLW Dataset
Results on WFLW dataset
| Model | Input Size | NME<sub>*test*</sub> | NME<sub>*pose*</sub> | NME<sub>*illumination*</sub> | NME<sub>*occlusion*</sub> | NME<sub>*blur*</sub> | NME<sub>*makeup*</sub> | NME<sub>*expression*</sub> | Details and Download |
| :-----: | :--------: | :------------------: | :------------------: | :--------------------------: | :-----------------------: | :------------------: | :--------------------: | :------------------------: | :--------------------: |
| HRNetv2-w18+Dark | 256x256 | 3.98 | 6.98 | 3.96 | 4.78 | 4.56 | 3.89 | 4.29 | [hrnetv2_dark_wflw.md](./wflw/hrnetv2_dark_wflw.md) |
| HRNetv2-w18+AWing | 256x256 | 4.02 | 6.94 | 3.97 | 4.78 | 4.59 | 3.87 | 4.28 | [hrnetv2_awing_wflw.md](./wflw/hrnetv2_awing_wflw.md) |
| HRNetv2-w18 | 256x256 | 4.06 | 6.97 | 3.99 | 4.83 | 4.58 | 3.94 | 4.33 | [hrnetv2_wflw.md](./wflw/hrnetv2_wflw.md) |