Body Pose Estimation on COCO-SinglePerson 1.0 (val)
83.5APRTMPose-l
Evaluation results
| Method | Links | |
|---|---|---|
| RTMPose-lBackbone=CSPNeXt-l, Detector=RTMDet-nano, Det. Input Size=320 × 320, Pose Input Size=256 x 192, GFLOPs=4.47, Extra Data=AIC(220K), Double inference=false2023.03 | 83.5 | |
| RTMPose-mBackbone=CSPNeXt-m, Detector=RTMDet-nano, Det. Input Size=320 × 320, Pose Input Size=256 x 192, GFLOPs=2.23, Extra Data=AIC(220K), Double inference=false2023.03 | 82.4 | |
| RTMPose-sBackbone=CSPNeXt-s, Detector=RTMDet-nano, Det. Input Size=320 × 320, Pose Input Size=256 x 192, GFLOPs=0.91, Extra Data=AIC(220K), Double inference=false2023.03 | 77.1 | |
| RTMPose-tBackbone=CSPNeXt-t, Detector=RTMDet-nano, Det. Input Size=320 × 320, Pose Input Size=256 x 192, GFLOPs=0.67, Extra Data=AIC(220K), Double inference=false2023.03 | 72.1 | |
| TinyPoseBackbone=Wider NLiteHRNet, Detector=PicoDet-s, Det. Input Size=320 × 320, Pose Input Size=256 x 192, GFLOPs=0.80, Extra Data=AIC(220K) +Internal(unknown), Double inference=false2023.03 | 69.4 | |
| MoveNet ThunderBackbone=MobileNetv2 depth×1.75, Detector=N/A, Det. Input Size=256 x 256, Pose Input Size=N/A, GFLOPs=2.44, Extra Data=Internal(23.5K), Double inference=true2023.03 | 64.8 | |
| TinyPoseBackbone=Wider NLiteHRNet, Detector=PicoDet-s, Det. Input Size=320 × 320, Pose Input Size=128 x 96, GFLOPs=0.55, Extra Data=AIC(220K), Double inference=false2023.03 | 58.6 | |
| MoveNet LightningBackbone=MobileNetv2, Detector=N/A, Det. Input Size=192 x 192, Pose Input Size=N/A, GFLOPs=0.54, Extra Data=Internal(23.5K), Double inference=true2023.03 | 53.6 | |
| BlazePose-FullBackbone=BlazePose, Detector=N/A, Det. Input Size=256 x 256, Pose Input Size=N/A, Extra Data=Internal(85K), Double inference=false2023.03 | 35.4 | |
| BlazePose-LiteBackbone=BlazePose, Detector=N/A, Det. Input Size=256 x 256, Pose Input Size=N/A, Extra Data=Internal(85K), Double inference=false2023.03 | 29.3 |