基于高精多尺度集成的轻量织物缺陷检测方法
|
|
张捷皓,张进峰,吴威涛,向忠
|
Lightweight fabric defect detection method based on high precision multi-scale integration
|
|
Jiehao ZHANG,Jinfeng ZHANG,Weitao WU,Zhong XIANG
|
|
| 表 4 不同模型基于ZY数据集的对比实验结果 |
| Tab.4 Comparative experimental results of different models based on ZY dataset |
|
| Method | P | R | F1 | mAP50 | mAP50:95 | Param/106 | GFOLPs/109 | FPS | | Faster RCNN | 0.432 | 0.813 | 0.56 | 0.634 | 0.250 | 137.0 | 370.3 | 11.2 | | SSD | 0.515 | 0.452 | 0.48 | 0.520 | 0.238 | 105.2 | 87.4 | 33.7 | | YOLOv5s | 0.965 | 0.790 | 0.87 | 0.880 | 0.624 | 7.0 | 16.0 | 52.9 | | YOLOv7 | 0.747 | 0.252 | 0.38 | 0.246 | 0.174 | 37.2 | 105.2 | 52.7 | | YOLOv8s | 0.945 | 0.801 | 0.87 | 0.886 | 0.623 | 11.1 | 28.7 | 51.3 | | YOLOv9s | 0.931 | 0.690 | 0.79 | 0.814 | 0.324 | 9.9 | 40.7 | 45.9 | | YOLOv10s | 0.864 | 0.867 | 0.87 | 0.879 | 0.335 | 8.0 | 24.5 | 63.5 | | YOLO11s | 0.722 | 0.636 | 0.68 | 0.688 | 0.399 | 9.4 | 21.3 | 66.7 | | CPD-YOLO | 0.974 | 0.848 | 0.91 | 0.912 | 0.640 | 5.3 | 13.3 | 51.0 |
|
|
|