Research Article
Object Detection Based on Swin Deformable Transformer-BiPAFPN-YOLOX
Table 3
Results of the ablation experiment of the two schemes on the COCO 2017 validation set.
| | Scheme A | Scheme B | AP (%) | AP50 | AP75 | APS | APM | APL | Param (M) | GFLOPs | Infer time (ms) | FPS |
| | | | 48.8 | 67.9 | 52.5 | 29.8 | 53.6 | 62.2 | 69.6 | 157.3 | 13.4 | 74.6 | | ✓ | | 49.3 | 68.2 | 52.8 | 30.5 | 54.1 | 62.7 | 62.8 | 148.7 | 11.7 | 85.5 | | | ✓ | 49.4 | 68.3 | 53.0 | 30.7 | 54.5 | 62.9 | 58.5 | 144.1 | 11.5 | 87.0 | | ✓ | ✓ | 49.7 | 68.6 | 53.2 | 31.2 | 54.9 | 63.3 | 44.5 | 110.7 | 10.1 | 98.7 |
|
|
