Table 4 Comparison and improvement results of backbone networks.
From: YOLO-STOD: an industrial conveyor belt tear detection model based on Yolov5 algorithm
Method | Accuracy (%) | Recall (%) | F1 (%) | Map (%) | FPS | GFLOPs/G | Params/M |
|---|---|---|---|---|---|---|---|
Yolov5s | 83.6 | 86.9 | 85.22 | 88.9 | 173.898 | 15.8 | 7.02 |
Yolov5s-BiFPN | 84 | 88.2 | 86.04 | 88.7 | 170.631 | 16.4 | 7.16 |
Yolov5s-CBAM | 83.8 | 86.4 | 85.08 | 89.9 | 155.329 | 15.9 | 7.06 |
Yolov5s-EfficientNet | 83.1 | 88.4 | 85.66 | 90.2 | 122.012 | 1.8 | 1.41 |
Yolov5s-DconV2 | 81.1 | 88.6 | 84.68 | 87.6 | 151.723 | 13.6 | 7.07 |
Yolov5s-C2f | 81.4 | 87.1 | 84.14 | 88.5 | 150.492 | 17.8 | 7.92 |
Yolov5s-FasterNet | 79 | 88.4 | 83.42 | 87.2 | 150.284 | 15.1 | 7.29 |
Yolov5s-SimAm | 80.1 | 90.7 | 85.08 | 88.2 | 172.746 | 15.8 | 7.01 |
Yolov5s-RepVGGNet | 78.3 | 84.5 | 81.28 | 85.3 | 170.566 | 6.1 | 1.87 |
Yolov5s-CotNet | 81 | 89.1 | 84.86 | 87.7 | 143.105 | 15.7 | 7.00 |
Yolov5s-BotNet | 83.6 | 91.3 | 87.28 | 90.8 | 176.575 | 20.7 | 7.68 |
