Simple Baselines for Human Pose Estimation and Tracking

Benchmark Model Rank Results
2d-human-pose-estimation-on-jhmdb-2d-posesSimplePose#3PCK: 94.4
2d-human-pose-estimation-on-ochumanResNet-152#5Test AP: 33.3Validation AP: 41.0
2d-human-pose-estimation-on-ochumanResNet-50#8Test AP: 30.4Validation AP: 37.8
keypoint-detection-on-cocoResNet-50#22Validation AP: 72.2
keypoint-detection-on-coco-test-challengeSimple Base+*#3AR: 80.5ARM: 75.3AP: 74.5AP50: 90.9AP75: 80.8APL: 87.5
keypoint-detection-on-coco-test-devSimple Base+*#2APM: 73.0APL: 82.7AP50: 92.4AP75: 84.0AR: 81.5AR50: 95.8
keypoint-detection-on-coco-test-devSimple Base#5APM: 70.3APL: 80.0AP50: 91.9AP75: 81.1AR: 79.0
keypoint-detection-on-ochumanResNet-152#4Test AP: 33.3Validation AP: 41.0
keypoint-detection-on-ochumanResNet-50#8Test AP: 29.5Validation AP: 32.1
multi-person-pose-estimation-on-crowdposeSimple baseline#20mAP @0.5:0.95: 60.8AP Easy: 71.4AP Medium: 61.2AP Hard: 51.2
multi-person-pose-estimation-on-ochumanSimplePose#7AP50: 37.4AP75: 26.8Validation AP: 24.1
pose-estimation-on-aicSimpleBaseline (ResNet-152)#8AP: 29.9AP75: 18.3AR: 34.3AR50: 76.9AP50: 73.8
pose-estimation-on-aicSimpleBaseline (ResNet-101)#9AP: 29.4AP75: 17.4AR: 33.7AR50: 76.3AP50: 73.6
pose-estimation-on-aicSimpleBaseline (ResNet-50)#10AP: 28.0AP75: 15.8AR: 32.1AR50: 74.1AP50: 71.6
pose-estimation-on-coco-test-devFlow-based (ResNet-152)#23AP: 73.7AP50: 91.9AP75: 81.1APL: 80APM: 70.3AR: 79
pose-estimation-on-coco-val2017SimpleBaseLine (256x192)#9AP: 70.4
pose-estimation-on-ochumanResNet-152#13Test AP: 33.3Validation AP: 41.0
pose-estimation-on-ochumanResNet-50#17Test AP: 29.5Validation AP: 32.1
pose-tracking-on-posetrack2017MSRA (FlowTrack)#3MOTA: 57.81mAP: 74.57
pose-tracking-on-posetrack2018MSRA#3MOTA: 61.37mAP: 74.03