mmpose/configs/body_2d_keypoint/topdown_regression

Top-down regression-based pose estimation

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, regression based methods directly regress the keypoint coordinates given the features extracted from the bounding box area, following the paradigm introduced in Deeppose: Human pose estimation via deep neural networks.

Results and Models

COCO Dataset

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-152+RLE	256x192	0.731	0.805	resnet_rle_coco.md
ResNet-101+RLE	256x192	0.722	0.768	resnet_rle_coco.md
ResNet-50+RLE	256x192	0.706	0.768	resnet_rle_coco.md
MobileNet-v2+RLE	256x192	0.593	0.644	mobilenetv2_rle_coco.md
ResNet-152	256x192	0.584	0.688	resnet_coco.md
ResNet-101	256x192	0.562	0.670	resnet_coco.md
ResNet-50	256x192	0.528	0.639	resnet_coco.md

MPII Dataset

Model	Input Size	PCKh@0.5	PCKh@0.1	Details and Download
ResNet-50+RLE	256x256	0.861	0.277	resnet_rle_mpii.md
ResNet-152	256x256	0.850	0.208	resnet_mpii.md
ResNet-101	256x256	0.841	0.200	resnet_mpii.md
ResNet-50	256x256	0.826	0.180	resnet_mpii.md