DJW cdece0b32a 第一次提交 | 10 місяців тому | |
---|---|---|
.. | ||
coco | 10 місяців тому | |
mpii | 10 місяців тому | |
README.md | 10 місяців тому |
Top-down methods divide the task into two stages: object detection, followed by single-object pose estimation given object bounding boxes. At the 2nd stage, SimCC based methods reformulate human pose estimation as two classification tasks for horizontal and vertical coordinates, and uniformly divide each pixel into several bins, thus obtain the keypoint coordinates given the features extracted from the bounding box area, following the paradigm introduced in SimCC: a Simple Coordinate Classification Perspective for Human Pose Estimation.
Results on COCO val2017 with detector having human AP of 56.4 on COCO val2017 dataset
Model | Input Size | AP | AR | Details and Download |
---|---|---|---|---|
ResNet-50+SimCC | 384x288 | 0.735 | 0.790 | resnet_coco.md |
ResNet-50+SimCC | 256x192 | 0.721 | 0.781 | resnet_coco.md |
S-ViPNAS-MobileNet-V3+SimCC | 256x192 | 0.695 | 0.755 | vipnas_coco.md |
MobileNet-V2+SimCC(wo/deconv) | 256x192 | 0.620 | 0.678 | mobilenetv2_coco.md |