Semi-supervised Human Pose Estimation in Art-historical Images

Benchmark Model Rank Results
semi-supervised-human-pose-estimation-onHRNet-W32#1mAP: 52.58mAP@0.5: 63.92mAP@0.75: 57.35
semi-supervised-human-pose-estimation-onHRNet-W32 (trained on PeopleArt)#2mAP: 29.71mAP@0.5: 36.37mAP@0.75: 32.72
semi-supervised-human-pose-estimation-onHRNet-W32 (trained on COCO 2017 with 0 % style-transferred Material)#3mAP: 25.25mAP@0.5: 31.73mAP@0.75: 28.10
semi-supervised-human-pose-estimation-onHRNet-W32 (trained on COCO 2017 with 100 % style-transferred Material)#4mAP: 25.18mAP@0.5: 31.67mAP@0.75: 28.13
semi-supervised-human-pose-estimation-onHRNet-W32 (trained on COCO 2017 with 50 % style-transferred Material)#5mAP: 24.13mAP@0.5: 30.52mAP@0.75: 26.65