Multi-Person Pose Estimation on COCO (test-dev)
79.2AP4xRSN-50
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 4xRSN-50Extra data=false, Pretrain=false, Backbone=4xRSN-50, Ensembled=true2020.03 | 79.2 | 94.4 | 87.1 | 76.1 | 83.8 | 84.1 | — | — | — | |
| 4xRSN-50Extra data=false, Pretrain=false, Backbone=4xRSN-50, Input Size=384×288, Params=111.8M, GFLOPS=65.9, Ensembled=false2020.03 | 78.6 | 94.3 | 86.6 | 75.5 | 83.3 | 83.8 | — | — | — | |
| 4xRSN-50Extra data=false, Pretrain=false, Backbone=4xRSN-50, Input Size=256x192, Params=111.8M, GFLOPS=29.3, Ensembled=false2020.03 | 78 | 94.2 | 86.5 | 75.3 | 82.2 | 83.4 | — | — | — | |
| MSPNExtra data=true, Pretrain=false, Backbone=4x ResNet-50, Input Size=384×288, Params=71.9M, GFLOPS=58.7, Ensembled=false2020.03 | 77.1 | 93.8 | 84.6 | 73.4 | 82.3 | 82.3 | — | — | — | |
| HRNet-W48Extra data=true, Backbone=HRNet-W48, Input Size=384×288, Params=63.6M, GFLOPS=32.9, Ensembled=false2020.03 | 77 | 92.7 | 84.5 | 73.4 | 83.1 | 82 | — | — | — | |
| SimpleBaseExtra data=true, Backbone=ResNet-152, Input Size=384×288, Ensembled=true2020.03 | 76.5 | 92.4 | 84 | 73 | 82.7 | 81.5 | — | — | — | |
| HRNet-W48Extra data=false, Backbone=HRNet-W48, Input Size=384×288, Params=63.6M, GFLOPS=32.9, Ensembled=false2020.03 | 75.5 | 92.5 | 83.3 | 71.9 | 81.5 | 80.5 | — | — | — | |
| 2xRSN-50Extra data=false, Pretrain=false, Backbone=2xRSN-50, Input Size=256×192, Params=54.0M, GFLOPS=13.9, Ensembled=false2020.03 | 75.5 | 93.6 | 84 | 73 | 79.6 | 81.3 | — | — | — | |
| HRNetBackbone=HRNet-w48, Multi-scale testing=false2020.06 | 75.5 | 92.5 | 83.3 | 71.9 | 81.5 | — | — | — | — | |
| RSNBackbone=2 × RSN-502021.07 | 75.5 | 93.6 | 84 | 73 | 79.6 | — | — | — | — | |
| HRNetPipeline=Two-Stage2022.12 | 75.5 | 92.5 | 83.3 | 71.9 | 81.5 | — | — | — | — | |
| HRNet-W32Extra data=false, Backbone=HRNet-W32, Input Size=384×288, Params=28.5M, GFLOPS=16, Ensembled=false2020.03 | 74.9 | 92.5 | 82.8 | 71.3 | 80.9 | 80.1 | — | — | — | |
| HRNetBackbone=HRNet-w322021.07 | 74.9 | 92.5 | 82.8 | 71.3 | 80.9 | — | — | — | — | |
| ECSIBackbone=ResNet-1522021.07 | 74.3 | 91.8 | 81.9 | 70.7 | 80.2 | — | — | — | — | |
| SimpleBaseExtra data=false, Backbone=ResNet-152, Input Size=384×288, Params=68.6M, GFLOPS=35.6, Ensembled=false2020.03 | 73.7 | 91.9 | 81.1 | 70.3 | 80 | 79 | — | — | — | |
| SimpleBaselinePipeline=Two-Stage2022.12 | 73.7 | 91.9 | 81.1 | 70.3 | 80 | — | — | — | — | |
| CPNExtra data=false, Backbone=ResNet-Inception, Input Size=384×288, Ensembled=true2020.03 | 73 | 91.7 | 80.9 | 69.5 | 78.1 | 79 | — | — | — | |
| RSN-50Extra data=false, Pretrain=false, Backbone=RSN-50, Input Size=256×192, Params=25.7M, GFLOPS=6.4, Ensembled=false2020.03 | 72.5 | 93 | 81.3 | 69.9 | 76.5 | 78.8 | — | — | — | |
| QueryPose-W48Pipeline=End-to-End, Backbone=HRNet-W482022.12 | 72.3 | 91.5 | 78.7 | 67.8 | 79 | — | — | — | — | |
| QueryPose-SwinLPipeline=End-to-End, Backbone=Swin-Large2022.12 | 72.2 | 92 | 78.8 | 67.3 | 79.4 | — | — | — | — | |
| CPNExtra data=false, Backbone=ResNet-Inception, Input Size=384×288, Params=58.8M, GFLOPS=29.2, Ensembled=false2020.03 | 72.1 | 91.4 | 80 | 68.7 | 77.2 | 78.5 | — | — | — | |
| CPNBackbone=ResNet-Inception2021.07 | 72.1 | 91.4 | 80 | 68.7 | 77.2 | — | — | — | — | |
| CPNPipeline=Two-Stage2022.12 | 72.1 | 91.4 | 80 | 68.7 | 77.2 | — | — | — | — | |
| LQRPipeline=End-to-End2022.12 | 71.7 | 90.4 | 78.7 | 67.3 | 78.5 | — | — | — | — | |
| RSN-18Extra data=false, Pretrain=false, Backbone=RSN-18, Input Size=256×192, Params=12.5M, GFLOPS=2.5, Ensembled=false2020.03 | 71.6 | 92.6 | 80.3 | 68.8 | 75.8 | 77.7 | — | — | — | |
| AdaptivePosePipeline=End-to-End2022.12 | 71.3 | 90 | 78.3 | 67.1 | 77.2 | — | — | — | — | |
| BAPoseInput Size=640, Testing Scale=Multi-Scale, Backbone=W482021.12 | 71.2 | 89.4 | 78.1 | 67.4 | 76.8 | 76.8 | 71.6 | 84 | — | |
| CenterGroupPipeline=End-to-End2022.12 | 71.1 | 90.5 | 77.5 | 66.9 | 76.7 | — | — | — | — | |
| HRNetInput Size=640, Testing Scale=Multi-Scale, Backbone=W482021.12 | 71 | 89.2 | 78 | 67.1 | 76.9 | 76.7 | 71.5 | 83.9 | — | |
| InsPoseBackbone=HRNet-w32, multi-scale testing=true2021.07 | 71 | 91.3 | 78 | 67.5 | 76.5 | — | — | — | — | |
| DEKRPipeline=End-to-End, Test-time augmentation=true2022.12 | 71 | 89.2 | 78 | 67.1 | 76.9 | — | — | — | — | |
| HigherHRNetBackbone=HRNet-w48, Multi-scale testing=true2020.06 | 70.5 | 89.3 | 77.2 | 66.6 | 75.8 | — | — | — | — | |
| HigherHRNetInput Size=640, Testing Scale=Multi-Scale, Backbone=W482021.12 | 70.5 | 89.3 | 77.2 | 66.6 | 75.8 | — | — | — | — | |
| HrHRNetPipeline=End-to-End2022.12 | 70.5 | 89.3 | 77.2 | 66.6 | 75.8 | — | — | — | — | |
| BAPoseInput Size=512, Testing Scale=Multi-Scale, Backbone=W322021.12 | 70.4 | 89.3 | 77.4 | 66 | 76.9 | 75.6 | 70.1 | 83.2 | — | |
| BAPoseInput Size=640, Testing Scale=Single-Scale, Backbone=W482021.12 | 70.3 | 89.6 | 77.5 | 65.9 | 77.1 | 75.4 | 69.8 | 83.2 | — | |
| SMPRBackbone=HRNet-w32, Multi-scale testing=true2020.06 | 70.2 | 89.7 | 77.5 | 65.9 | 77.2 | — | — | — | — | |
| HRNetInput Size=640, Testing Scale=Single-Scale, Backbone=W482021.12 | 70 | 89.4 | 77.3 | 65.7 | 76.9 | 75.4 | 69.7 | 83.2 | — | |
| HRNetInput Size=512, Testing Scale=Multi-Scale, Backbone=W322021.12 | 69.6 | 89 | 76.6 | 65.2 | 76.5 | 75.1 | 69.5 | 82.8 | — | |
| InsPoseBackbone=HRNet-w322021.07 | 69.3 | 90.3 | 76 | 64.8 | 76.1 | — | — | — | 0.15 | |
| Point-Set AnchorsBackbone=HRNet-w48, Multi-scale testing=true2020.06 | 68.7 | 89.9 | 76.3 | 64.8 | 75.3 | — | — | — | — | |
| Point-set AnchorsInput Size=640, Testing Scale=Multi-Scale2021.12 | 68.7 | 89.9 | 76.3 | 64.8 | 75.3 | 74.8 | 69.6 | 82.1 | — | |
| PersonLabInput Size=1401, Testing Scale=Multi-Scale2021.12 | 68.7 | 89 | 75.4 | 64.1 | 75.5 | 75.4 | 69.7 | 83 | — | |
| Point-Set AnchorsBackbone=HRNet-w48, multi-scale testing=true2021.07 | 68.7 | 89.9 | 76.3 | 64.8 | 75.3 | — | — | — | 0.27 | |
| PointSetNetPipeline=End-to-End2022.12 | 68.7 | 89.9 | 76.3 | 64.8 | 75.3 | — | — | — | — | |
| G-RMIExtra data=true, Backbone=ResNet-101, Input Size=353×257, Params=42.6M, GFLOPS=57, Ensembled=false2020.03 | 68.5 | 87.1 | 75.5 | 65.8 | 73.3 | 73.3 | — | — | — | |
| HigherHRNetInput Size=640, Testing Scale=Single-Scale, Backbone=W482021.12 | 68.4 | 88.2 | 75.1 | 64.4 | 74.2 | — | — | — | — | |
| SMPRBackbone=HRNet-w32, Multi-scale testing=false2020.06 | 68.2 | 88.7 | 75.3 | 63.3 | 75.4 | — | — | — | — | |
| BAPoseInput Size=512, Testing Scale=Single-Scale, Backbone=W322021.12 | 68 | 88 | 74.8 | 62.4 | 76.6 | 72.9 | 66.1 | 82.4 | — | |
| InsPoseBackbone=ResNet-101, multi-scale testing=true2021.07 | 67.8 | 90.3 | 74.9 | 64.3 | 73.4 | — | — | — | — | |
| Integral PosePipeline=Two-Stage2022.12 | 67.8 | 88.2 | 74.8 | 63.9 | 74 | — | — | — | — | |
| HGGInput Size=512, Testing Scale=Multi-Scale2021.12 | 67.6 | 85.1 | 73.7 | 62.7 | 74.6 | 71.3 | — | — | — | |
| DGCNBackbone=ResNet-1522021.07 | 67.4 | 88 | 74.4 | 63.6 | 73 | — | — | — | — | |
| HRNetInput Size=512, Testing Scale=Single-Scale, Backbone=W322021.12 | 67.3 | 87.9 | 74.1 | 61.5 | 76.1 | 72.4 | 65.4 | 81.9 | — | |
| SMPRBackbone=ResNet-50, Multi-scale testing=true2020.06 | 67.1 | 88.5 | 74.2 | 62 | 74.7 | — | — | — | — | |
| InsPoseBackbone=ResNet-50, multi-scale testing=true2021.07 | 67.1 | 89.7 | 74 | 62.6 | 73.8 | — | — | — | — | |
| SPMBackbone=Hourglass-8 stacked, Multi-scale testing=false2020.06 | 66.9 | 88.5 | 72.9 | 62.6 | 73.1 | — | — | — | — | |
| SPMTesting Scale=Single-Scale2021.12 | 66.9 | 88.5 | 72.9 | 62.6 | 73.1 | — | — | — | — | |
| SPMBackbone=Hourglass-8 stacked, multi-scale testing=true, additional refinement=true2021.07 | 66.9 | 88.5 | 72.9 | 62.6 | 73.1 | — | — | — | — | |
| SPMPipeline=End-to-End, Single-person refinement=true, Test-time augmentation=true2022.12 | 66.9 | 88.5 | 72.9 | 62.6 | 73.1 | — | — | — | — | |
| PifPafTesting Scale=Single-Scale2021.12 | 66.7 | — | — | 62.4 | 72.9 | — | — | — | — | |
| PifPafBackbone=ResNet-1522021.07 | 66.7 | — | — | 62.4 | 72.9 | — | — | — | 0.26 | |
| PersonLabBackbone=ResNet-152, Multi-scale testing=false2020.06 | 66.5 | 88 | 72.6 | 62.4 | 72.3 | — | — | — | — | |
| PersonLabInput Size=1401, Testing Scale=Single-Scale2021.12 | 66.5 | 88.9 | 72.6 | 62.4 | 72.3 | 71 | 66.1 | 77.7 | — | |
| HigherHRNetBackbone=HRNet-w32, Multi-scale testing=false2020.06 | 66.4 | 87.5 | 72.8 | 61.2 | 74.2 | — | — | — | — | |
| HigherHRNetBackbone=HRNet-w322021.07 | 66.4 | 87.5 | 72.8 | 61.2 | 74.2 | — | — | — | 0.6 | |
| InsPoseBackbone=ResNet-1012021.07 | 66.3 | 89.2 | 73 | 61.2 | 73.9 | — | — | — | 0.1 | |
| GOG (Ours final)Backbone=Hourglass-104, Input=768, Inference scale=single-scale2021.07 | 65.6 | — | — | 63.3 | 68.8 | — | — | — | — | |
| AEInput Size=512, Testing Scale=Multi-Scale2021.12 | 65.5 | 86.8 | 72.3 | 60.6 | 72.6 | 70.2 | 64.6 | 78.1 | — | |
| AEBackbone=Hourglass-4 stacked, multi-scale testing=true, additional refinement=true2021.07 | 65.5 | 87.1 | 71.4 | 61.3 | 71.5 | — | — | — | 0.52 | |
| PersonLabBackbone=ResNet-1012021.07 | 65.5 | 87.1 | 71.4 | 61.3 | 71.5 | — | — | — | — | |
| AEPipeline=End-to-End, Single-person refinement=true, Test-time augmentation=true2022.12 | 65.5 | 86.8 | 72.3 | 60.6 | 72.6 | — | — | — | — | |
| PersonLabBackbone=ResNet-101, Input=1401, Inference scale=single-scale2021.07 | 65.5 | — | — | 61.3 | 71.5 | — | — | — | — | |
| InsPoseBackbone=ResNet-502021.07 | 65.4 | 88.9 | 71.7 | 60.2 | 72.7 | — | — | — | 0.08 | |
| G-RMIExtra data=false, Backbone=ResNet-101, Input Size=353×257, Params=42.6M, GFLOPS=57, Ensembled=false2020.03 | 64.9 | 85.5 | 71.3 | 62.3 | 70 | 69.7 | — | — | — | |
| G-RMIPipeline=Two-Stage2022.12 | 64.9 | 85.5 | 71.3 | 62.3 | 70 | — | — | — | — | |
| PIFPAFBackbone=ResNet-101, Input=641, Inference scale=single-scale2021.07 | 64.9 | — | — | 60.6 | 71.2 | — | — | — | — | |
| SMPRBackbone=ResNet-50, Multi-scale testing=false2020.06 | 64.7 | 87.3 | 71.2 | 59.1 | 72.7 | — | — | — | — | |
| GOG (Ours final)Backbone=Hourglass-104, Input=640, Inference scale=single-scale2021.07 | 64.7 | — | — | 60.7 | 70.4 | — | — | — | — | |
| Simple PoseBackbone=IMHourglass, Input=768, Inference scale=single-scale2021.07 | 64.6 | — | — | 60.3 | 71.6 | — | — | — | — | |
| Bottom-up HRNet-W32Input=>640, Inference scale=single-scale2021.07 | 64.1 | — | — | 57.4 | 73.9 | — | — | — | — | |
| DirectPoseBackbone=ResNet-1012021.07 | 63.3 | 86.7 | 69.4 | 57.8 | 71.2 | — | — | — | — | |
| AEBackbone=Hourglass-4 stacked, Multi-scale testing=true2020.06 | 63 | 85.7 | 68.9 | 58 | 70.4 | — | — | — | — | |
| DirectPoseBackbone=ResNet-50, Multi-scale testing=true2020.06 | 63 | 86.8 | 69.3 | 59.1 | 69.3 | — | — | — | — | |
| CenterNetBackbone=Hourglass-2 stacked (104), Multi-scale testing=false2020.06 | 63 | 86.8 | 69.6 | 58.9 | 70.4 | — | — | — | — | |
| CenterNet-HGInput Size=512, Testing Scale=Single-Scale2021.12 | 63 | 86.8 | 69.6 | 58.9 | 70.4 | — | — | — | — | |
| CenterNetBackbone=Hourglass-1042021.07 | 63 | 86.8 | 69.6 | 58.9 | 70.4 | — | — | — | 0.16 | |
| CenterNet-HGPipeline=End-to-End2022.12 | 63 | 86.8 | 69.6 | 58.9 | 70.4 | — | — | — | — | |
| CenterNetBackbone=Hourglass-104, Input=~640, Inference scale=single-scale2021.07 | 63 | — | — | 58.9 | 70.4 | — | — | — | — | |
| MDN3Testing Scale=Single-Scale2021.12 | 62.9 | 85.1 | 69.4 | 58.8 | 71.4 | — | — | — | — | |
| Mask-RCNNBackbone=ResNet-50, Multi-scale testing=false2020.06 | 62.7 | 87 | 68.4 | 57.4 | 71.1 | — | — | — | — | |
| Mask-RCNNBackbone=ResNet-502021.07 | 62.7 | 87 | 68.4 | 57.4 | 71.1 | — | — | — | 0.08 | |
| DirectPoseBackbone=ResNet-50, Multi-scale testing=false2020.06 | 62.2 | 86.4 | 68.2 | 56.7 | 69.8 | — | — | — | — | |
| DirectPoseBackbone=ResNet-502021.07 | 62.2 | 86.4 | 68.2 | 56.7 | 69.8 | — | — | — | — | |
| CMUposeExtra data=false, Ensembled=false2020.03 | 61.8 | 84.9 | 67.5 | 57.1 | 68.2 | 66.5 | — | — | — | |
| CMU-PoseBackbone=3CM-3PAF (102), Multi-scale testing=false2020.06 | 61.8 | 84.9 | 67.5 | 57.1 | 68.2 | — | — | — | — | |
| OpenPoseTesting Scale=Single-Scale2021.12 | 61.8 | 84.9 | 67.5 | 57.1 | 68.2 | 66.5 | — | — | — | |
| CMU-PoseBackbone=3CM-3PAF (102), multi-scale testing=true, additional refinement=true2021.07 | 61.8 | 84.9 | 67.5 | 57.1 | 68.2 | — | — | — | — | |
| CMU-PosePipeline=End-to-End, Single-person refinement=true2022.12 | 61.8 | 84.9 | 67.5 | 57.1 | 68.2 | — | — | — | — | |
| CenterNet-DLAInput Size=512, Testing Scale=Single-Scale2021.12 | 57.9 | 84.7 | 63.1 | 52.5 | 67.4 | — | — | — | — |