Table 4 Comparison with other methods on the MSCOCO validation set.
From: Curvelet-enhanced transformer architecture for blurred action fine-grained detection
Methods | Backbone | Frame size | Parameters (M) | AP | AP50 | AP75 |
|---|---|---|---|---|---|---|
TransPose | TransPose-H-A4 | 256 × 192 | 17.3 | 0.753 | – | – |
SimCC | ResNet-50 | 256 × 192 | 25.7 | 0.708 | – | – |
HRNet | HRNet-W32 | 256 × 192 | 28.5 | 0.734 | 0.895 | 0.807 |
PRTR | ResNet-50 | 384 × 288 | 41.5 | 0.682 | 0.882 | 0.752 |
EBA | ResNet-18 | 256 × 256 | 17.0 | 0.713 | 0.915 | 0.781 |
RIFormer | HRFormer-B | 256 × 192 | 43.2 | 0.756 | 0.908 | 0.828 |
BCIR | ResNet-50 | 256 × 192 | 34.0 | 0.675 | 0.872 | 0.740 |
AECA | ResNet-18 | 384 × 288 | 19.0 | 0.745 | 0.925 | 0.814 |
MCTN | DETR | 256 × 192 | 24.5 | 0.759 | 0.926 | 0.822 |
MCTN | RT-DETRv3 | 256 × 192 | 32.8 | 0.767 | 0.938 | 0.836 |
MCTN | DETR | 384 × 288 | 34.6 | 0.761 | 0.922 | 0.828 |
MCTN | RT-DETRv3 | 384 × 288 | 40.4 | 0.766 | 0.941 | 0.833 |
