Real-time Fabric Defect Detection Algorithm Based on S-YOLOV3 Model

Fig. 6. Samples of YID fabric defects. (a) Broken filament 1; (b) broken filament 2; (c) broken filament 3; (d) broken filament 4; (e) hole; (f) broken-picks; (g) cracked-ends; (h) pulp-stain; (i) filaments; (j) stain; (k) driving-defects; (l) wrinkle

Download full size

Fig. 7. Test results for each defect in FID dataset

Download full size

Fig. 8. Test results for each defect in YID dataset

Download full size

Fig. 9. Test results of fabric defects

Download full size

Fig. 10. FID defect detection results. (a) Broken-end; (b) hole; (c) netting-multiple; (d) thick-bar; (e) thin-bar

Download full size

Fig. 11. YID defect detection results. (a) Broken-filament 1; (b) broken-filament 2; (c) broken-filament 3; (d) wrinkle; (e) broken-picks; (f) stain; (g) pulp-stain; (h) driving-defects

Download full size

Table 1. Performance comparison of different detection frameworks on test sets
View table
Table 1. Performance comparison of different detection frameworks on test sets
Framework FID YID
mAP /% FPS mAP /% FPS
Mobilenet1.0 81 45 80 48
Darknet53 93 41 95 40

Table 2. Performance comparison of different models on test set
View table
Table 2. Performance comparison of different models on test set
Model mAP /% FPS
GAN+Faster R-CNN 92.0 5
YOLOV2 76.8 67
YOLOV3 94.0 40

Table 3. Comparison of test results between YOLOV3 and S-YOLOV3
View table
Table 3. Comparison of test results between YOLOV3 and S-YOLOV3
Model Number ofarguments /M mAP /% FPS
YOLOV3 65 94 40
S-YOLOV3 15 94 55

Tools

Get Citation

Copy Citation Text

Jun Zhou, Junfeng Jing, Huanhuan Zhang, Zhen Wang, Hanlin Huang. Real-time Fabric Defect Detection Algorithm Based on S-YOLOV3 Model[J]. Laser & Optoelectronics Progress, 2020, 57(16): 161001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites