2D Human Pose Estimation on COCO 2017 (val)
79.3APPCT
Evaluation results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PCTBackbone=Swin-Huge, Input size=256 x 256, GFLOPs=118.2, Speed (fps)=31.72023.03 | 79.3 | 91.5 | 85.9 | — | — | — | — | — | — | — | — | — | — | — | — | |
| ReChannel-9BModel Size=9B2026.07 | 79.2 | 92.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViTPoseBackbone=ViT-Huge, Input size=256 x 192, GFLOPs=122.9, Speed (fps)=21.82023.03 | 79.1 | 91.6 | 85.7 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CCNet + RLEConfidence=Reg, Backbone=HRNet-W48, Input Size=384×288, #Params (M)=75.6, #GFLOPS=33.32023.11 | 78.8 | 92.6 | 85.1 | 80.6 | — | 77 | 82.9 | — | — | — | — | — | — | — | — | |
| ViTPoseBackbone=ViT-Large, Input size=256 x 192, GFLOPs=59.8, Speed (fps)=40.52023.03 | 78.3 | 91.4 | 85.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCTBackbone=Swin-Large, Input size=256 x 256, GFLOPs=34.1, Speed (fps)=76.42023.03 | 78.3 | 91.4 | 85.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| RTMPose-lFramework=MMPose, Input Size=256 x 192, GFLOPs=4.16, Deployment Hardware=Snapdragon 865, Deployment Library=ncnn2023.03 | 78.3 | — | — | — | — | — | — | — | — | — | — | — | — | 85.75 | 45.37 | |
| CCNet + HRNetConfidence=Hm, Backbone=HRNet-W48, Input Size=384×288, #Params (M)=63.69, #GFLOPS=15.362023.11 | 78.3 | 93.6 | 85.1 | 80.9 | — | 75.5 | 83.4 | — | — | — | — | — | — | — | — | |
| ViTPose-LVariant=L2026.07 | 78.3 | 91.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CCNet + ViTPoseConfidence=Hm, Backbone=ViT-Base, Input Size=256×192, #Params (M)=90.07, #GFLOPS=17.912023.11 | 78.1 | 93.7 | 85 | 80.4 | — | 75.4 | 83.3 | — | — | — | — | — | — | — | — | |
| ReChannel-4BModel Size=4B2026.07 | 78 | 91.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RLEConfidence=Reg, Backbone=HRNet-W48, Input Size=384×288, #Params (M)=75.6, #GFLOPS=33.32023.11 | 77.9 | 92.4 | 84.5 | 80.6 | — | 77.1 | 81.4 | — | — | — | — | — | — | — | — | |
| DPIT-LBackbone=HRNet-W48-s, Input Size=256 x 192, #Params=38.0M2022.09 | 77.8 | 93.6 | 84.8 | 80.3 | — | 74.8 | 82.2 | — | — | — | — | — | — | — | — | |
| UDPBackbone=HRNet-W48, Input size=384 x 288, GFLOPs=35.5, Speed (fps)=67.92023.03 | 77.8 | 92 | 84.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| PCTBackbone=Swin-Base, Input size=256 x 256, GFLOPs=15.2, Speed (fps)=115.12023.03 | 77.7 | 91.2 | 84.7 | — | — | — | — | — | — | — | — | — | — | — | — | |
| CCNet + PoseurConfidence=Reg, Backbone=ResNet-50, Input Size=256×192, #Params (M)=33.1, #GFLOPS=4.62023.11 | 77.7 | 92.7 | 84.2 | 79.7 | — | 74.9 | 82.3 | — | — | — | — | — | — | — | — | |
| EvoPose2D-L*PT (Pre-training)=N, Input Size=512x384, Params (M)=14.7, FLOPs (G)=17.7, Size (MB)=60.6, FPS (GPU)=15.9, Post-processing=PoseFix2020.11 | 77.5 | 90.9 | 83.6 | 82.5 | — | 74 | 84.2 | — | — | — | — | — | — | — | — | |
| CCNet + RLEConfidence=Reg, Backbone=HRNet-W32, Input Size=256×192, #Params (M)=39.3, #GFLOPS=7.12023.11 | 77.5 | 92.6 | 84.2 | 79.4 | — | 75.9 | 81.3 | — | — | — | — | — | — | — | — | |
| HRNetConfidence=Hm, Backbone=HRNet-W48, Input Size=384×288, #Params (M)=63.62, #GFLOPS=15.312023.11 | 77.4 | 93.4 | 84.4 | 80.9 | — | 74.8 | 82.1 | — | — | — | — | — | — | — | — | |
| UniFormer-BArch.=CNN+Trans, Input Size=448x320, #Param(M)=53.5, FLOPs(G)=29.6, Pre-training=ImageNet-1K [22]2022.01 | 77.4 | 91.1 | 84.4 | 82.5 | — | 70.2 | 80.6 | — | — | — | — | — | — | — | — | |
| MogaNet-BType=CNN, Crop size=384 x 288, Params (M)=47, FLOPs (G)=24.42022.11 | 77.3 | 91.4 | 84 | 82.2 | — | — | — | — | — | — | — | — | — | — | — | |
| MogaNet-BDate=Ours, Crop size=384 x 288, Param. (M)=47.4, FLOPs (G)=24.42022.11 | 77.3 | 91.4 | 84 | 82.2 | — | — | — | — | — | — | — | — | — | — | — | |
| HRNet-W48*PT (Pre-training)=Y, Input Size=384x288, Params (M)=63.6, FLOPs (G)=35.3, Size (MB)=259, FPS (GPU)=16.2, Post-processing=PoseFix2020.11 | 77.3 | 90.9 | 83.5 | 82 | — | 73.5 | 84.4 | — | — | — | — | — | — | — | — | |
| I2R-NetInput size=384x288, #Params=43.7(+0.5)M, FLOPs=29.6(+2.8)G, 1st stage=HRFormer-B+2022.06 | 77.3 | 91 | 83.6 | 82.1 | — | 73 | 84.5 | — | — | 77.7 | 88.6 | — | — | — | — | |
| RTMPose-mFramework=MMPose, Input Size=256 x 192, GFLOPs=1.93, Deployment Hardware=Snapdragon 865, Deployment Library=ncnn2023.03 | 77.3 | — | — | — | — | — | — | — | — | — | — | — | — | 49.46 | 26.44 | |
| CCNet + SBLConfidence=Hm, Backbone=ResNet-152, Input Size=384×288, #Params (M)=68.71, #GFLOPS=12.832023.11 | 77.3 | 93.5 | 84.1 | 79.3 | — | 74 | 81.6 | — | — | — | — | — | — | — | — | |
| ViTPoseConfidence=Hm, Backbone=ViT-Base, Input Size=256×192, #Params (M)=89.99, #GFLOPS=17.852023.11 | 77.3 | 93.5 | 84.5 | 80.4 | — | 75 | 81.6 | — | — | — | — | — | — | — | — | |
| HRFormer-BType=Transformer, Crop size=384 x 288, Params (M)=54, FLOPs (G)=30.72022.11 | 77.2 | 91 | 83.6 | 82 | — | — | — | — | — | — | — | — | — | — | — | |
| HRFormerBackbone=HRFormer-B, Input size=384 x 288, GFLOPs=29.1, Speed (fps)=25.22023.03 | 77.2 | 91 | 83.6 | — | — | — | — | — | — | — | — | — | — | — | — | |
| HRFormer-BArch.=CNN+Trans, Input Size=384x288, #Param(M)=43.2, FLOPs(G)=30.7, Pre-training=ImageNet-1K [22]2022.01 | 77.2 | 91 | 83.6 | 82 | — | 73.2 | 84.2 | — | — | — | — | — | — | — | — | |
| HRFormer-B*Input size=384x288, #Params=43.2M, FLOPs=26.8G2022.06 | 77.1 | 90.9 | 83.5 | 82 | — | 69.8 | 80.5 | — | — | 77.7 | 88.3 | — | — | — | — | |
| HRNetBackbone=H-W48, Input Size=256 × 192, #Params=63.6M2022.09 | 77.1 | 93.6 | 84.7 | 79.9 | — | 74.1 | 81.9 | — | — | — | — | — | — | — | — | |
| CCNet + RLEConfidence=Reg, Backbone=ResNet-152, Input Size=384×288, #Params (M)=58.3, #GFLOPS=11.32023.11 | 77.1 | 92.6 | 83.2 | 79.2 | — | 75.6 | 81.3 | — | — | — | — | — | — | — | — | |
| ScaleNASBackbone=ScaleNet-P4, Input size=384x288, #Params=64.3M, GFLOPS=32.62020.11 | 77 | 90.9 | 83.6 | 81.8 | — | 73 | 84.2 | — | — | — | — | — | — | — | — | |
| CCNet + HRNetConfidence=Hm, Backbone=HRNet-W32, Input Size=256×192, #Params (M)=28.62, #GFLOPS=7.762023.11 | 77 | 93.7 | 84 | 79.3 | — | 74 | 81 | — | — | — | — | — | — | — | — | |
| DPIT-BBackbone=HRNet-W32-s, Input Size=256 x 192, #Params=20.8M2022.09 | 76.9 | 93.5 | 83.7 | 79.6 | — | 73.7 | 81.5 | — | — | — | — | — | — | — | — | |
| SimCCBackbone=HRNet-W48, Input size=384 x 288, GFLOPs=32.9, Speed (fps)=71.42023.03 | 76.9 | 90.9 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DARKBackbone=HRNet-W48, Input size=384 x 288, GFLOPs=35.5, Speed (fps)=62.12023.03 | 76.8 | 90.6 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| PoseurConfidence=Reg, Backbone=ResNet-50, Input Size=256×192, #Params (M)=33.1, #GFLOPS=4.62023.11 | 76.8 | 92.6 | 83.7 | 79.7 | — | 74.2 | 81.4 | — | — | — | — | — | — | — | — | |
| Uniformer-BType=Hybrid, Crop size=384 x 288, Params (M)=54, FLOPs (G)=14.82022.11 | 76.7 | 90.8 | 84 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| Uniformer-BDate=ICLR'2022, Crop size=384 x 288, Param. (M)=53.5, FLOPs (G)=14.82022.11 | 76.7 | 90.8 | 84 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| RLEConfidence=Reg, Backbone=HRNet-W32, Input Size=256×192, #Params (M)=39.3, #GFLOPS=7.12023.11 | 76.7 | 92.4 | 83.5 | 79.4 | — | 76 | 79.3 | — | — | — | — | — | — | — | — | |
| UniFormer-BArch.=CNN+Trans, Input Size=384x288, #Param(M)=53.5, FLOPs(G)=22.1, Pre-training=ImageNet-1K [22]2022.01 | 76.7 | 90.8 | 84 | 81.9 | — | 69.3 | 79.7 | — | — | — | — | — | — | — | — | |
| ProbPose-sBounding Box=GT2024.12 | 76.6 | — | — | — | 76.4 | — | — | — | — | — | — | — | — | — | — | |
| EvoPose2D-LPT (Pre-training)=N, Input Size=512x384, Params (M)=14.7, FLOPs (G)=17.7, Size (MB)=60.6, FPS (GPU)=15.92020.11 | 76.6 | 90.5 | 83 | 82.3 | — | 72.7 | 83.4 | — | — | — | — | — | — | — | — | |
| Simple BaselineBackbone=R-152, Input Size=384 × 288, #Params=68.6M2022.09 | 76.6 | 92.6 | 83.6 | 79.3 | — | 73.7 | 81.3 | — | — | — | — | — | — | — | — | |
| SimBa.Backbone=Swin-Base, Input size=256 x 256, GFLOPs=16.6, Speed (fps)=74.42023.03 | 76.6 | 91.4 | 84.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| RTMPose-lBackbone=CSPNeXt-l, Detector=RTMDet-m, Det. Input Size=640x640, Pose Input Size=256x192, GFLOPs=4.16, Extra Data=AIC(220K), Framework=MMPose [11]2023.03 | 76.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HRNetBackbone=H-W32, Input Size=256 × 192, #Params=28.5M2022.09 | 76.5 | 93.5 | 83.7 | 79.3 | — | 73.9 | 80.8 | — | — | — | — | — | — | — | — | |
| SBLConfidence=Hm, Backbone=ResNet-152, Input Size=384×288, #Params (M)=68.64, #GFLOPS=12.772023.11 | 76.5 | 92.5 | 83.6 | 79.3 | — | 73.6 | 81.2 | — | — | — | — | — | — | — | — | |
| MogaNet-SType=CNN, Crop size=384 x 288, Params (M)=29, FLOPs (G)=13.52022.11 | 76.4 | 91 | 83.3 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| MogaNet-SDate=Ours, Crop size=384 x 288, Param. (M)=29.0, FLOPs (G)=13.52022.11 | 76.4 | 91 | 83.3 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| I2R-NetInput size=256x192, #Params=43.7(+0.5)M, FLOPs=12.8(+0.6)G, 1st stage=HRFormer-B*2022.06 | 76.4 | 90.8 | 83.2 | 81.4 | — | 72.3 | 83.7 | — | — | 76.9 | 88.1 | — | — | — | — | |
| HRNet-W48Type=CNN, Crop size=384 x 288, Params (M)=64, FLOPs (G)=32.92022.11 | 76.3 | 90.8 | 82 | 81.2 | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-LType=Transformer, Crop size=384 x 288, Params (M)=203, FLOPs (G)=86.92022.11 | 76.3 | 91.2 | 83 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| HRNet-W48Date=CVPR 2019, Crop size=384 x 288, Param. (M)=63.6, FLOPs (G)=32.92022.11 | 76.3 | 90.8 | 82 | 81.2 | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-LDate=ICCV'2021, Crop size=384 x 288, Param. (M)=203.4, FLOPs (G)=86.92022.11 | 76.3 | 91.2 | 83 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| HRNet-W48Backbone=HRNet-W48, Pretrain=Y, Input size=384 x 288, #Params=63.6M, GFLOPS=32.92019.02 | 76.3 | 90.8 | 82.9 | 81.2 | — | 72.3 | 83.4 | — | — | — | — | — | — | — | — | |
| HRNetV1Backbone=HRNetV1-W48, Pretrain=true, Input size=384 x 288, #Params=63.6M, GFLOPS=32.92019.08 | 76.3 | 90.8 | 82.9 | 81.2 | — | 72.3 | 83.4 | — | — | — | — | — | — | — | — | |
| HRNetBackbone=HRNet-W48, Input size=384x288, #Params=63.6M, GFLOPS=32.92020.11 | 76.3 | 90.8 | 82.9 | 81.2 | — | 72.3 | 83.4 | — | — | — | — | — | — | — | — | |
| ScaleNASBackbone=ScaleNet-P3, Input size=384x288, #Params=26.2M, GFLOPS=14.32020.11 | 76.3 | 90.7 | 82.9 | 81.3 | — | 72.5 | 83.3 | — | — | — | — | — | — | — | — | |
| HRNet-W48PT (Pre-training)=Y, Input Size=384x288, Params (M)=63.6, FLOPs (G)=35.3, Size (MB)=259, FPS (GPU)=16.22020.11 | 76.3 | 90.8 | 82.9 | 81.2 | — | 72.3 | 83.4 | — | — | — | — | — | — | — | — | |
| HRNetBackbone=HRNet-W482021.09 | 76.3 | 90.8 | 82.9 | — | — | 72.3 | 83.4 | — | — | — | — | — | — | — | — | |
| HRNetBackbone=HRNet-W48, Input size=384 x 288, GFLOPs=35.5, Speed (fps)=75.52023.03 | 76.3 | 90.8 | 82.9 | — | — | — | — | — | — | — | — | — | — | — | — | |
| RTMPose-lBackbone=CSPNeXt-l, Detector=Faster-RCNN, Det. Input Size=N/A, Pose Input Size=256x192, GFLOPs=4.16, Framework=MMPose [11]2023.03 | 76.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RTMPose-lInput Size=256 x 192, GFLOPS=4.16, CPU(ms)=18.847, GPU(ms)=3.459, Flip test=not used2023.03 | 76.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RLEConfidence=Reg, Backbone=ResNet-152, Input Size=384×288, #Params (M)=58.3, #GFLOPS=11.32023.11 | 76.3 | 92.4 | 82.6 | 79.2 | — | 75.6 | 79.7 | — | — | — | — | — | — | — | — | |
| HRNet-W48Arch.=CNN, Input Size=384x288, #Param(M)=63.6, FLOPs(G)=32.9, Pre-training=ImageNet-1K [22]2022.01 | 76.3 | 90.8 | 82.9 | 81.2 | — | 72.3 | 83.4 | — | — | — | — | — | — | — | — | |
| HRNet-W48Backbone=HRNet-W482026.07 | 76.3 | 90.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ProbPose-s-DHBounding Box=GT, Double Heatmap=true, Wider Field of View=true2024.12 | 76.2 | — | — | — | 75.4 | — | — | — | — | — | — | — | — | — | — | |
| UniFormer-SArch.=CNN+Trans, Input Size=448x320, #Param(M)=25.2, FLOPs(G)=14.8, Pre-training=ImageNet-1K [22]2022.01 | 76.2 | 90.6 | 83.2 | 81.4 | — | 68.6 | 79.4 | — | — | — | — | — | — | — | — | |
| HRNetConfidence=Hm, Backbone=HRNet-W32, Input Size=256×192, #Params (M)=28.54, #GFLOPS=7.72023.11 | 76 | 93.5 | 83.4 | 79.3 | — | 73.7 | 80 | — | — | — | — | — | — | — | — | |
| Uniformer-SType=Hybrid, Crop size=384 x 288, Params (M)=25, FLOPs (G)=11.12022.11 | 75.9 | 90.6 | 83.4 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt-BType=CNN, Crop size=384 x 288, Params (M)=94, FLOPs (G)=36.62022.11 | 75.9 | 90.6 | 83.1 | 81.1 | — | — | — | — | — | — | — | — | — | — | — | |
| Uniformer-SDate=ICLR'2022, Crop size=384 x 288, Param. (M)=25.2, FLOPs (G)=11.12022.11 | 75.9 | 90.6 | 83.4 | 81.4 | — | — | — | — | — | — | — | — | — | — | — | |
| ViTPose-sBounding Box=GT2024.12 | 75.9 | — | — | — | 75.3 | — | — | — | — | — | — | — | — | — | — | |
| MSPNPT (Pre-training)=Y, Input Size=256x192, Params (M)=120, FLOPs (G)=19.92020.11 | 75.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniFormer-SArch.=CNN+Trans, Input Size=384x288, #Param(M)=25.2, FLOPs(G)=11.1, Pre-training=ImageNet-1K [22]2022.01 | 75.9 | 90.6 | 83.4 | 81.4 | — | 68.6 | 79 | — | — | — | — | — | — | — | — | |
| HRNet-W32Type=CNN, Crop size=384 x 288, Params (M)=29, FLOPs (G)=162022.11 | 75.8 | 90.6 | 82.7 | 81 | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt-SType=CNN, Crop size=384 x 288, Params (M)=55, FLOPs (G)=21.82022.11 | 75.8 | 90.7 | 83.1 | 81 | — | — | — | — | — | — | — | — | — | — | — | |
| HRNet-W32Backbone=HRNet-W32, Pretrain=Y, Input size=384 x 288, #Params=28.5M, GFLOPS=16.02019.02 | 75.8 | 90.6 | 82.7 | 81 | — | 71.9 | 82.8 | — | — | — | — | — | — | — | — | |
| HRNetV1Backbone=HRNetV1-W32, Pretrain=true, Input size=384 x 288, #Params=28.5M, GFLOPS=162019.08 | 75.8 | 90.6 | 82.7 | 81 | — | 71.9 | 82.8 | — | — | — | — | — | — | — | — | |
| HRNet-W32PT (Pre-training)=Y, Input Size=384x288, Params (M)=28.5, FLOPs (G)=16, Size (MB)=119, FPS (GPU)=22.72020.11 | 75.8 | 90.6 | 82.7 | 81 | — | 71.9 | 82.8 | — | — | — | — | — | — | — | — | |
| TransPose-HInput size=256x192, #Params=17.5M, FLOPs=21.8G2022.06 | 75.8 | 90.1 | 82.1 | 80.8 | — | 71.9 | 82.8 | — | — | 76.4 | 87.2 | — | — | — | — | |
| I2R-NetInput size=256x192, #Params=18.0(+0.5)M, FLOPs=22.7(+0.9)G, 1st stage=TransPose-H2022.06 | 75.8 | 90.4 | 82.1 | 80.9 | — | 72 | 82.9 | — | — | 76.6 | 87.3 | — | — | — | — | |
| ED-PoseParadigm=One-stage, Backbone=Swin-L, Loss=R., pre-trained=Objects3652023.02 | 75.8 | 92.3 | 82.9 | — | — | 70.4 | 83.5 | — | — | — | — | 142 | — | — | — | |
| TransPoseBackbone=HRNet-W48, Input size=256 x 192, GFLOPs=21.8, Speed (fps)=56.72023.03 | 75.8 | 90.1 | 82.1 | — | — | — | — | — | — | — | — | — | — | — | — | |
| TokenPoseBackbone=HRNet-W48, Input size=256 x 192, GFLOPs=22.1, Speed (fps)=52.92023.03 | 75.8 | 90.3 | 82.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViTPoseBackbone=ViT-Base, Input size=256 x 192, GFLOPs=17.9, Speed (fps)=113.52023.03 | 75.8 | 90.7 | 83.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| TransPose-H-A6Arch.=CNN+Trans, Input Size=256x192, #Param(M)=17.5, FLOPs(G)=21.8, Pre-training=ImageNet-1K [22]2022.01 | 75.8 | — | — | 80.8 | — | — | — | — | — | — | — | — | — | — | — | |
| TokenPose-L/D24Arch.=CNN+Trans, Input Size=256x192, #Param(M)=27.5, FLOPs(G)=11.0, Pre-training=ImageNet-1K [22]2022.01 | 75.8 | 90.3 | 82.5 | 80.9 | — | 72.3 | 82.7 | — | — | — | — | — | — | — | — | |
| HRNet-W32Arch.=CNN, Input Size=384x288, #Param(M)=28.5, FLOPs(G)=16.0, Pre-training=ImageNet-1K [22]2022.01 | 75.8 | 90.6 | 82.7 | 81 | — | 71.9 | 82.8 | — | — | — | — | — | — | — | — | |
| TokenPose-LVariant=L2026.07 | 75.8 | 90.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MogaNet-TType=CNN, Crop size=384 x 288, Params (M)=8, FLOPs (G)=4.92022.11 | 75.7 | 90.6 | 82.6 | 80.9 | — | — | — | — | — | — | — | — | — | — | — | |
| RTMPose-lBackbone=CSPNeXt-l, Detector=YOLOv3, Det. Input Size=608x608, Pose Input Size=256x192, GFLOPs=4.16, Framework=AlphaPose [18] (Fair Comparison)2023.03 | 75.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RTMPose-mBackbone=CSPNeXt-m, Detector=RTMDet-m, Det. Input Size=640x640, Pose Input Size=256x192, GFLOPs=1.93, Extra Data=AIC(220K), Framework=MMPose [11]2023.03 | 75.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HRFormer-B*Input size=256x192, #Params=43.2M, FLOPs=12.2G2022.06 | 75.6 | 90.8 | 82.8 | 80.8 | — | 71.7 | 82.6 | — | — | 76.5 | 87.2 | — | — | — | — | |
| HRFormer-BArch.=CNN+Trans, Input Size=256x192, #Param(M)=43.2, FLOPs(G)=14.1, Pre-training=ImageNet-1K [22]2022.01 | 75.6 | 90.8 | 82.8 | 80.8 | — | 71.7 | 82.6 | — | — | — | — | — | — | — | — | |
| HRFormer-SArch.=CNN+Trans, Input Size=384x288, #Param(M)=7.8, FLOPs(G)=7.2, Pre-training=ImageNet-1K [22]2022.01 | 75.6 | 90.3 | 82.2 | 80.7 | — | 71.6 | 82.5 | — | — | — | — | — | — | — | — | |
| Simple BaselineBackbone=R-101, Input Size=384 x 288, #Params=53.0M2022.09 | 75.5 | 92.5 | 82.6 | 78.4 | — | 72.4 | 80.8 | — | — | — | — | — | — | — | — |