Table 2 Experimental results of TT-100 K dataset in mainstream model.
From: Small traffic sign recognition method based on improved YOLOv7
Algorithm | Accuracy (P) (%) | Recall rate (R) (%) | mAP@0.5(%) | mAP@0.5:0.95(%) | FPS |
|---|---|---|---|---|---|
Deformable DETR43 | 47.40 | 65.60 | 62.90 | 41.20 | – |
DINO44 | 61.80 | 73.50 | 76.00 | 53.90 | – |
DAMO-YOLO48 | 69.70 | 72.30 | 78.60 | 59.50 | 38.9 |
YOLOv324 | 83.84 | 73.87 | 81.28 | 62.97 | 25.3 |
YOLOv5 | 70.82 | 62.19 | 67.70 | 51.78 | 68.5 |
YOLOv626 | 85.70 | 76.90 | 83.30 | 65.30 | 75.6 |
YOLOv712 | 87.31 | 83.15 | 89.26 | 70.38 | 38.4 |
RFB Net45 | 86.92 | 83.05 | 85.25 | 74.44 | – |
ScratchDet46 | 84.57 | 77.16 | 82.46 | 74.02 | – |
CAB Net47 | 88.01 | 84.71 | 88.09 | 78.08 | – |
RTSRA11 | 87.65 | 82.23 | 90.35 | 71.12 | – |
Ours | 88.69 | 88.41 | 92.74 | 72.67 | 42.2 |
