Keypoint Detection on MS COCO 2017 (test-dev)
79.2APRSN
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| RSNBackbone=4xRSN-50, Input size=384 × 288, External data=false, Ensemble=true2020.02 | 79.2 | 94.4 | 87.1 | 76.1 | 83.8 | 84.1 | — | — | — | — | |
| CCMBackbone=HRNet-w48, Input size=384 × 288, External data=true, Ensemble=true2020.02 | 78.9 | 93.8 | 86 | 75 | 84.5 | 83.6 | — | — | — | — | |
| RSNBackbone=4xRSN-50, #Params=111.8M, GFLOPS=65.9, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 78.6 | 94.3 | 86.6 | 75.5 | 83.3 | 83.8 | — | — | — | — | |
| MSPNBackbone=4xResNet-50, Input size=384 × 288, External data=true, Ensemble=true2020.02 | 78.1 | 94.1 | 85.9 | 74.5 | 83.3 | 83.1 | — | — | — | — | |
| CCMBackbone=HRNet-w48, #Params=63.7M, GFLOPS=36.6, Input size=384 × 288, External data=true, Ensemble=false2020.02 | 78 | 93.4 | 85.1 | 74 | 83.6 | 83 | — | — | — | — | |
| DARKBackbone=HRNet-w48, #Params=63.6M, GFLOPS=32.9, Input size=384 × 288, External data=true, Ensemble=false2020.02 | 77.4 | 92.6 | 84.6 | 73.6 | 83.7 | 82.3 | — | — | — | — | |
| CCMBackbone=ResNet-152, #Params=63.5M, GFLOPS=40.1, Input size=384 × 288, External data=true, Ensemble=false2020.02 | 77.3 | 93 | 84.8 | 73.3 | 83.1 | 82.3 | — | — | — | — | |
| MSPNBackbone=4xResNet-50, #Params=71.9M, GFLOPS=58.7, Input size=384 × 288, External data=true, Ensemble=false2020.02 | 77.1 | 93.8 | 84.6 | 73.4 | 82.3 | 82.3 | — | — | — | — | |
| HRNetBackbone=HRNet-w48, #Params=63.6M, GFLOPS=35.4, Input size=384 × 288, External data=true, Ensemble=false2020.02 | 77 | 92.7 | 84.5 | 73.4 | 83.1 | 82 | — | — | — | — | |
| CCMBackbone=HRNet-w48, #Params=63.7M, GFLOPS=36.6, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 76.6 | 92.8 | 84.1 | 72.6 | 82.4 | 81.7 | — | — | — | — | |
| BaselineBackbone=ResNet-152, Input size=384 × 288, External data=true, Ensemble=true2020.02 | 76.5 | 92.4 | 84 | 73 | 82.7 | 81.5 | — | — | — | — | |
| DARKBackbone=HRNet-w48, #Params=63.6M, GFLOPS=32.9, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 76.2 | 92.5 | 83.6 | 72.5 | 82.4 | 81.1 | — | — | — | — | |
| MSPNBackbone=4xResNet-50, #Params=71.9M, GFLOPS=58.7, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 76.1 | 93.4 | 83.8 | 72.3 | 81.5 | 81.6 | — | — | — | — | |
| CCMBackbone=ResNet-152, #Params=63.5M, GFLOPS=40.1, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 75.8 | 92.7 | 83.4 | 71.8 | 81.5 | 80.9 | — | — | — | — | |
| HRNet-W48*Backbone=HRNet-W48, Pretrain=true, Train Input=384x288, Test Input=384x2882019.11 | 75.5 | — | — | 71.9 | 81.5 | 80.5 | 95.7 | — | — | — | |
| HRNetBackbone=HRNet-w48, #Params=63.6M, GFLOPS=35.4, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 75.5 | 92.5 | 83.3 | 71.9 | 81.5 | 80.5 | — | — | — | — | |
| BaselineBackbone=ResNet-152, #Params=68.6M, GFLOPS=35.9, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 73.7 | 91.9 | 81.1 | 70.3 | 80 | 79 | — | — | — | — | |
| CPN+backbone=ResNet-Inception [36], ensembled=true2017.11 | 73 | 91.7 | 80.9 | 69.5 | 78.1 | 79 | 95.1 | 85.9 | 74.8 | 84.7 | |
| CPNBackbone=ResNet-Inception, Input size=384 × 288, External data=false, Ensemble=true2020.02 | 73 | 91.7 | 80.9 | 69.5 | 78.1 | 79 | — | — | — | — | |
| RMPEBackbone=Hourglass, #Params=28.1M, GFLOPS=36.7, Input size=320 × 256, External data=false, Ensemble=false2020.02 | 72.3 | 89.2 | 79.1 | 68 | 78.6 | — | — | — | — | — | |
| CPNbackbone=ResNet-Inception [36], ensembled=false2017.11 | 72.1 | 91.4 | 80 | 68.7 | 77.2 | 78.5 | 95.1 | 85.3 | 74.2 | 84.3 | |
| CPN*Backbone=ResNet-Inception, Pretrain=true, Train Input=384x288, Test Input=384x2882019.11 | 72.1 | — | — | 68.7 | 77.2 | 78.5 | 95.1 | — | — | — | |
| CPNBackbone=ResNet-Inception, Input size=384 × 288, External data=false, Ensemble=false2020.02 | 72.1 | 91.4 | 80 | 68.7 | 77.2 | 78.5 | — | — | — | — | |
| G-RMIextra data=true2017.11 | 68.5 | 87.1 | 75.5 | 65.8 | 73.3 | 73.3 | 90.1 | 79.5 | 68.1 | 80.4 | |
| G-RMIBackbone=ResNet-101, #Params=42.6M, GFLOPS=57, Input size=353 × 257, External data=true, Ensemble=false2020.02 | 68.5 | 87.1 | 75.5 | 65.8 | 73.3 | 73.3 | — | — | — | — | |
| Ours-4 (final), w/ FLBackbone=4-stage IMHN plus, Pretrain=false, Train Input=512x512, Test Input=~512^2, Refine=false2019.11 | 68.1 | — | — | 66.8 | 70.5 | 72.1 | 88.2 | — | — | — | |
| PersonLab*Backbone=ResNet-101, Pretrain=true, Train Input=801x801, Test Input=~1401^2, Refine=false2019.11 | 67.8 | — | — | 63 | 74.8 | 74.5 | 92.2 | — | — | — | |
| Ours-5, w/ FLBackbone=4-stage IMHN plus, Pretrain=false, Train Input=512x512, Test Input=~384^2, Refine=false2019.11 | 67.6 | — | — | 64.5 | 72.6 | 71.3 | 87.6 | — | — | — | |
| PifPaf*Backbone=ResNet-152, Pretrain=true, Train Input=401x401, Test Input=~641^2, Refine=false2019.11 | 66.7 | — | — | 62.4 | 72.9 | 72.2 | 90.9 | — | — | — | |
| Ours-3, w/ FLBackbone=4-stage IMHN, Pretrain=false, Train Input=384x384, Test Input=~384^2, Refine=false2019.11 | 66.2 | — | — | 66.4 | 66.6 | 71.2 | 88.6 | — | — | — | |
| Associative Embedding2017.11 | 65.5 | 86.8 | 72.3 | 60.6 | 72.6 | 70.2 | 89.5 | 76 | 64.6 | 78.1 | |
| AE*Backbone=4-stage Hourglass, Pretrain=false, Train Input=512x512, Test Input=~512^2, Refine=true2019.11 | 65.5 | — | — | 60.6 | 72.6 | 70.2 | 89.5 | — | — | — | |
| Ours-2, w/ FLBackbone=3-stage IMHN, Pretrain=false, Train Input=384x384, Test Input=~384^2, Refine=false2019.11 | 65.2 | — | — | 63.7 | 68.5 | 69.8 | 87.7 | — | — | — | |
| G-RMIextra data=false2017.11 | 64.9 | 85.5 | 71.3 | 62.3 | 70 | 69.7 | 88.7 | 75.5 | 64.4 | 77.1 | |
| PifPafBackbone=ResNet-101, Pretrain=true, Train Input=401x401, Test Input=~641^2, Refine=false2019.11 | 64.9 | — | — | 60.6 | 71.2 | 70.3 | 90.2 | — | — | — | |
| G-RMI*Backbone=ResNet-101, Pretrain=true, Train Input=353x257, Test Input=353x2572019.11 | 64.9 | — | — | 62.3 | 70 | 69.7 | 88.7 | — | — | — | |
| G-RMIBackbone=ResNet-101, #Params=42.6M, GFLOPS=57, Input size=353 × 257, External data=false, Ensemble=false2020.02 | 64.9 | 85.5 | 71.3 | 62.3 | 70 | 69.7 | — | — | — | — | |
| Mask-RCNN2017.11 | 63.1 | 87.3 | 68.7 | 57.8 | 71.4 | — | — | — | — | — | |
| Mask-RCNNBackbone=ResNet-50, External data=false, Ensemble=false2020.02 | 63.1 | 87.3 | 68.7 | 57.8 | 71.4 | — | — | — | — | — | |
| CMU-Pose2017.11 | 61.8 | 84.9 | 67.5 | 57.1 | 68.2 | 66.5 | 87.2 | 71.8 | 60.6 | 74.6 | |
| CMU-Pose*Backbone=6-stage CMU-Net, Pretrain=false, Train Input=368x368, Test Input=~368^2, Refine=true2019.11 | 61.8 | — | — | 57.1 | 68.2 | 66.5 | 87.2 | — | — | — | |
| Ours-1, w/ FLBackbone=3-stage CMU-Net, Pretrain=false, Train Input=368x368, Test Input=~368^2, Refine=false2019.11 | 59.3 | — | — | 56.2 | 63.8 | 63.5 | 84.6 | — | — | — | |
| CMU-Pose (baseline)Backbone=6-stage CMU-Net, Pretrain=false, Train Input=368x368, Test Input=~368^2, Refine=false2019.11 | 52.9 | — | — | 50.9 | 57.2 | 57 | 79.2 | — | — | — |