Table 2 Comparison of detection performance with mainstream and state-of-the-art algorithms.
From: Balancing complexity and accuracy for defect detection on filters with an improved RT-DETR
Model | Params(M) | FLOPs(G) | Precision/% | Recall/(%) | mAP@0.5(%) | mAP@0.5:0.95(%) | FPS |
|---|---|---|---|---|---|---|---|
Faster R-CNN | 42.0 | 180.0 | 90.70 | 90.21 | 90.02 | 65.84 | 31.0 |
SSD | 21.1 | 62.7 | 85.54 | 82.59 | 82.30 | 56.58 | 120.2 |
YOLOv5m | 21.2 | 49.0 | 86.08 | 88.48 | 88.26 | 64.33 | 166.7 |
YOLOv7-s AF | 11.0 | 28.1 | 87.66 | 87.27 | 86.95 | 60.32 | 197.6 |
YOLOv8m | 25.9 | 78.9 | 90.02 | 91.18 | 90.62 | 68.47 | 117.4 |
YOLOv9 | 25.3 | 92.87 | 92.15 | 91.18 | 92.40 | 68.07 | 74.9 |
YOLOv10 | 19.1 | 92.0 | 91.16 | 89.65 | 90.57 | 63.57 | 149.8 |
DETR | 41.2 | 107.43 | 87.17 | 85.22 | 84.37 | 60.21 | 47.9 |
RT-DETR | 20.1 | 58.29 | 83.20 | 90.54 | 90.32 | 65.37 | 134.1 |
RT-DETRv2 | 20.0 | 59.1 | 90.27 | 90.25 | 90.24 | 65.93 | 136.5 |
RT-DETR-LGA(ours) | 18.7 | 50.67 | 98.54 | 97.07 | 97.67 | 75.72 | 118.1 |
