Table 1.
Comparison of variant models and our model.
| Method | mAP | FPS | Batch size | Input resolution |
|---|---|---|---|---|
| FasterRCNN | 0.5396 | 7 | 2 | 600 × 600 |
| MaskRCNN | 0.6493 | 9 | 2 | 600 × 600 |
| EfficientDet | 0.6520 | 107 | 8 | 512 × 512 |
| YOLOv3 | 0.5806 | 23 | 2 | 608 × 608 |
| YOLOv4 | 0.6385 | 127 | 2 | 608 × 608 |
| YOLOv5 | 0.6674 | 151 | 2 | 608 × 608 |
| SSD300 | 0.6465 | 45 | 2 | 300 × 300 |
| SSD300 | 0.6460 | 62 | 8 | 300 × 300 |
| SSD512 | 0.6477 | 19 | 2 | 512 × 512 |
| SSD512 | 0.6478 | 21 | 8 | 512 × 512 |
| Our model 512 | 0.6825 | 147 | 2 | 512 × 512 |
| Our model 512 | 0.6893 | 151 | 8 | 512 × 512 |
| Our model 1024 | 0.6756 | 79 | 2 | 1024 × 1024 |
The bold values were used to label the best performing score compared with others.