Chinese Journal of Lasers, Volume. 52, Issue 12, 1202105(2025)
A Lightweight Real‐Time Weld Defect Classification Model
Figures & Tables(11)
Fig. 1. Schematic diagrams of models. (a) YOLOv8n-cls model; (b) TDRE-YOLO-cls model
Fig. 3. Schematic diagrams of RepConv. (a)Training multi-branch structure;(b)convolutional layer and batch normalization layer fusion;(c)inference single convolution
Fig. 6. Sample set examples. (a) Burr defect; (b) dimpling defect; (c) pore defect ; (d) no obvious defect
Table 1. Comparison of results of classification models on validation set
View tableTable 1. Comparison of results of classification models on validation set
Model Top-1 /
%
Weighted
precision /%
Weighted
recall /%
YOLOv8n-cls 82.5 82.522 82.776 YOLOv11n-cls 82.1 81.998 82.245 YOLOv8s-cls 83.4 83.443 83.929 YOLOv11s-cls 82.3 82.261 82.752 TDRE-YOLO-cls 84.8 84.760 84.885
Table 2. Comparison of results of classification models on test set
View tableTable 2. Comparison of results of classification models on test set
Model Top-1 /% Weighted precision /% Weighted recall /% Time /ms Size/kB YOLOv8n-cls 81.6 81.604 81.758 0.9 2907 YOLOv11n-cls 81.9 81.867 82.223 0.9 3124 YOLOv8s-cls 81.7 81.736 82.091 1.4 10028 YOLOv11s-cls 82.1 82.131 82.558 1.4 10784 Mobilenetv3 82.4 82.370 82.689 3.9 6078 Shufflenetv2 80.9 80.944 81.108 4.4 5086 TDRE-YOLO-cls 84.0 83.972 83.998 0.9 1393
Table 3. Comparison of ablation experimental results
View tableTable 3. Comparison of ablation experimental results
Model Top-1 /% Weighted precision /% Weighted recall /% Time /ms Size/kB Model_NoRCR 82.8 82.790 83.159 0.9 1373 Model_NoSPP 81.5 81.472 81.908 1.0 1704 Model_NoDSPA 81.5 81.474 81.982 1.0 2570 TDR-YOLO-cls 83.6 83.576 82.833 0.9 1385 TDRE-YOLO-cls 84.0 83.972 83.998 0.9 1393
Table 4. Precisions of classification models on four samples
View tableTable 4. Precisions of classification models on four samples
Model Precision /% Burr defect Dimpling defect Pore defect No obvious defectYOLOv8n-cls 81.646 80.000 83.077 82.069 YOLOv11n-cls 84.688 74.556 83.077 83.098 YOLOv8s-cls 81.957 81.481 80.769 82.428 YOLOv11s-cls 86.250 72.189 84.615 82.394 TDRE-YOLO-cls 85.535 81.928 83.969 82.877
Tools
Get Citation
Copy Citation Text
Yunhao Li, Chengtie Li, Qiuming Li. A Lightweight Real‐Time Weld Defect Classification Model[J]. Chinese Journal of Lasers, 2025, 52(12): 1202105
Paper Information
Category: Laser Forming Manufacturing
Received: Nov. 26, 2024
Accepted: Mar. 6, 2025
Published Online: May. 28, 2025
The Author Email: Yunhao Li (872047087@qq.com), Qiuming Li (2272268@stu.neu.edu.cn)
CSTR:32183.14.CJL241390
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20