Table 3 Comparisons on the COCO validation set.
From: An efficient and accurate 2D human pose estimation method using VTTransPose network
Method | Input size | AP | AR | #Params (M) | FLOPs (G) |
|---|---|---|---|---|---|
SimpleBaseline-Res5027 | 256 × 192 | 70.4 | 76.3 | 34.0 | 8.9 |
SimpleBaseline-Res10127 | 256 × 192 | 71.4 | 76.3 | 53.0 | 12.4 |
SimpleBaseline-Res15227 | 256 × 192 | 72.0 | 77.8 | 68.6 | 35.3 |
TransPose-R-A3*17 | 256 × 192 | 71.5 | 76.9 | 5.0 | 5.4 |
TransPose-R-A317 | 256 × 192 | 71.7 | 77.1 | 5.2 | 8.0 |
TransPose-R-A417 | 256 × 192 | 72.6 | 78.0 | 6.0 | 8.9 |
HRNet-W328 | 256 × 192 | 74.4 | 79.8 | 28.5 | 7.2 |
HRNet-W488 | 256 × 192 | 75.1 | 80.4 | 63.6 | 14.6 |
TokenPose-B16 | 256 × 192 | 74.7 | 80.0 | 13.5 | 5.7 |
DistilPose-S18 | 256 × 192 | 71.6 | – | 5.4 | 2.38 |
DistilPose-L18 | 256 × 192 | 74.4 | – | 21.3 | 10.33 |
GTPose-B19 | 256 × 192 | 75.0 | 80.1 | 13.5 | – |
TransPose-H-A417 | 256 × 192 | 75.3 | 80.3 | 17.3 | 17.5 |
TransPose-H-A617 | 256 × 192 | 75.8 | 80.8 | 17.5 | 21.8 |
TransPose-H–S17 | 256 × 192 | 74.2 | 78.0 | 8.0 | 10.2 |
VTTranspose (ours) | 256 × 192 | 74.6 | 78.5 | 6.0 | 5.4 |