Explicit Box Detection Unifies End-to-End Multi-Person Pose Estimation

Open paper
Benchmark Model Rank Results
2d-human-pose-estimation-on-human-artED-Pose (R50)#2AP: 0.723AP (gt bbox): /
human-pose-keypoint-estimation-on-coco-test-devED-Pose (Swin-L)#46AP: 72.7AP50: 92.3AP75: 80.9
human-pose-keypoint-estimation-on-coco-test-devED-Pose (ResNet-50)#74AP: 69.8AP50: 90.2AP75: 77.2
human-pose-keypoint-estimation-on-coco-val2017ED-Pose (Swin-L, Objects365 pretrain)#47AP: 75.8AP50: 92.3AP75: 82.9
human-pose-keypoint-estimation-on-coco-val2017ED-Pose (Swin-L)#62AP: 74.3AP50: 91.5AP75: 81.6
human-pose-keypoint-estimation-on-coco-val2017ED-Pose (ResNet-50)#79AP: 71.6AP50: 89.6AP75: 78.1
multi-person-pose-estimation-on-crowdposeED-Pose (Swin-L)#4mAP @0.5:0.95: 76.6AP Easy: 83.0AP Medium: 77.3AP Hard: 68.3

Rank counts only results with a code link.