Metal Surface Defect Detection and Classification using EfficientNetV2 and YOLOv5

다국어 초록 (Multilingual Abstract)

Detection and classification of steel surface defects are critical for product quality control in the steel industry. However, due to its low accuracy and slow speed, the traditional approach cannot be effectively used in a production line. The current, widely used algorithm (based on deeplearning) has an accuracy problem, and there are still rooms for development. This paper proposes a method of steel surface defect detection combining EfficientNetV2 for image classification and YOLOv5 as an object detector. Shorter training time and high accuracy are advantages of this model. Firstly, the image input into EfficientNetV2 model classifies defect classes and predicts probability of having defects. If the probability of having a defect is less than 0.25, the algorithm directly recognizes that the sample has no defects. Otherwise, the samples are further input into YOLOv5 to accomplish the defect detection process on the metal surface. Experiments show that proposed model has good performance on the NEU dataset with an accuracy of 98.3%. Simultaneously, the average training speed is shorter than other models.

참고문헌 (Reference)

1 박미소 ; 윤홍주 ; 김나경 ; 김보람, "적조 탐지를 위한 기계학습 모델 비교 연구" 한국전자통신학회 16 (16): 1363-1371, 2021

2 Z. Ge, "Yolox : Exceeding yolo series in 2021"

3 김민수 ; 문미경 ; 한창희, "YOLO와 OCR 알고리즘에 기반한 시각 장애우를 위한 유통기한 알림 시스템" 한국전자통신학회 16 (16): 1329-1338, 2021

4 이군호 ; 문미경, "YOLO알고리즘을 활용한 시각장애인용 식사보조 시스템 개발" 한국전자통신학회 16 (16): 1001-1010, 2021

5 H. Zhang, "Varifocalnet : An iou-aware dense object detector" 8514-8523, 2021

6 G. Jocher, "Ultralytics/yolov5: v3.1 – Bug Fixes and Performance Improvements"

7 M. Everingham, "The Pascal Visual Object Classes Challenge: A Retrospective" 111 : 98-136, 2014

8 C. M. Wang, "Surface quality detection of cold-rolled strip based on BP neural network" 6 : 106-108, 2007

9 Y. Yang, "Surface defect detection of steel strip based on CNN" 2019 : 25-29, 2019

10 Ch. Wang, "Scaled-yolov4 : Scaling cross stage partial network" 13029-13038, 2021

11 H. Wang, "Research on surface defect detection of metal sheet and strip based on multi-level feature FasterR-CNN" 2021 (2021): 262-269, 2021

12 Z. Yang, "Reppoints : Point set representation for object detection" 9657-9666, 2019

13 A. Paszke, "Pytorch: An imperative style, high-performance deep learning library" 8026-8037, 2019

14 X. Zhou, "Probabilistic two-stage detection"

15 K. Wang, "Panet: Few-shot image semantic segmentation with prototype alignment" 9197-9206, 2019

16 X. Zhou, "Objects as points"

17 T. Y. Lin, "Microsoft coco: Common objects in context" 740-755, 2014

18 Y. Xu, "Metal Surface Defect Detection Using Modified YOLO" 14 : 257-, 2021

19 M. Versaci, "Innovative fuzzy techniques for characterizing defects inultrasonic non-destructive evaluation" 2014 : 201-232, 2014

20 M. Vannocci, "Flatness defect detection and classification in hot rolled steel strips using convolutional neural networks" 11507 : 220-234, 2019

21 Z. Tian, "Fcos : Fully convolutional one-stage object detection" 9627-9636, 2019

22 Sh. Ren, "Faster R-CNN:Towards real-time object detection with region proposal networks" 28 : 91-99, 2015

23 R. Wei, "Enhanced faster Region Convolutional Neural Networks for Steel Surface Defect Detection" 60 (60): 539-545, 2020

24 M. Tan, "EfficientNetV2: Smaller Models and Faster Training" PMLR 139 : 10096-10106, 2021

25 M. Tan, "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks" PMLR 97 : 6105-6114, 2020

26 D. He, "Design of multi-scale receptive field convolutional neural network for surface inspection of hot-rolled steels" 89 : 12-20, 2019

27 X. Lv, "Deep metallic surface defect detection : the new benchmark and detection network" 20 (20): 1562-, 2020

28 C. Y. Wang, "CSPNet: A new backbone that can enhance learning capability of CNN" 390-391, 2020

29 S. Wang, "Automatic Detection and Classification of Steel Surface Defect Using Deep Convolutional Neural Networks" 11 (11): 388-, 2021

30 Y. He, "An end-to-end steel surface defect detection approach via fusing multiple hierarchical features" 69 (69): 1493-1504, 2020

31 K. Song, "A noise robust method based on completed localbinary patterns for hot-rolled steel strip surface defects" 285 (285): 858-864, 2013

32 R. Xu, "A Forest Fire Detection System Based on Ensemble Learning" 12 (12): 217-, 2021