Bottom-up Human Pose Estimation on COCO 2017 (val)
70.4APScaleNAS
Evaluation results
| Method | Links | |
|---|---|---|
| ScaleNASBackbone=ScaleNet-P4(Ours), Input size=640, #Params=64.4M, GFLOPS=141.5, Multi-scale test=false2020.11 | 70.4 | |
| HigherHRNetBackbone=HRNet-W48, Input size=640, #Params=63.8M, GFLOPS=154.3, Multi-scale test=false2020.11 | 69.9 | |
| ScaleNASBackbone=ScaleNet-P1(Ours), Input size=512, #Params=28.6M, GFLOPS=46.9, Multi-scale test=false2020.11 | 67.8 | |
| HigherHRNetBackbone=HRNet-W32, Input size=512, #Params=28.6M, GFLOPS=47.9, Multi-scale test=false2020.11 | 67.1 |