An augmentation method of defect detection dataset based on M-DCGAN

TANG Luyuan, ZHAO Hong, WANG Ning, HAN Bing, WANG Yuanyuan, LI Wangyang

doi:10.16411/j.cnki.issn1006-7736.2023.02.016

Journal of Dalian Maritime University >

2023 , Vol. 49 >Issue 2: 148 - 160

DOI: https://doi.org/10.16411/j.cnki.issn1006-7736.2023.02.016

An augmentation method of defect detection dataset based on M-DCGAN

Expand

(1a. College of Marine Electrical Engineering; b. Marine Engineering College, Dalian Maritime University, Dalian 116026, China; 2. Shanghai Ship and Shipping Research Institute Co., Ltd, Shanghai 200137, China)

Online published: 2023-03-07

Fold

Abstract

Aiming at the insufficient surface defect dataset in intelligent manufacturing, an augmentation method based on mask deep convolution generative adversarial networks(M-DCGAN) was proposed. Firstly, the U-Net-like structure was built by adding an upsampling module to the discriminator and enhancing network depth of the discriminator and generator,a binarization mask extraction method of defect position was designed based on Canny edge detection. Then, a defect target position attention guidance mechanism was established by defining the loss function image mask dependent,the spectral normalization layer and dropout layer were inserted into the network to enhance training stability and maintain the generated images diversity. The experimental results on the strip steel defect dataset show that the images quality generated by the proposed model is higher than that of DCGAN, WGAN-GP, and InfoGAN. Enrich training data by the proposed M-DCGAN algorithm can significantly improve and surpass the defect detection accuracy of traditional augmentation algorithms in eight classic methods such as YOLOv5, SSD, Faster R-CNN, and YOLOv3, which verifies the effectiveness of the proposed algorithm.

Key words： defect detection; augmentation of dataset; deep learning; generative adversarial network; M-DCGAN（Mask Deep Convolution Generative Adversarial Networks)

Cite this article

TANG Luyuan, ZHAO Hong, WANG Ning, HAN Bing, WANG Yuanyuan, LI Wangyang . An augmentation method of defect detection dataset based on M-DCGAN[J]. Journal of Dalian Maritime University, 2023 , 49(2) : 148 -160 . DOI: 10.16411/j.cnki.issn1006-7736.2023.02.016

References

[1]DESHPANDE A M, MINAI A A, KUMAR M. One-shot recognition of manufacturing defects in steel surfaces[J]. Procedia Manufacturing, 2020, 48: 1064-1071.

[2]WANG N, WANG Y, ER M J. Review on deep learning techniques for marine object recognition: architectures and algorithms[J]. Control Engineering Practice, 2022, 118: 104458.

[3]SHORTEN C, KHOSHGOFTAAR T M. A survey on image data augmentation for deep learning[J]. Journal of Big Data, 2019, 6(1):1-48.

[4]CHEN T, WANG N, WANG R, et al. One-stage CNN detector-based benthonic organisms detection with limited training dataset[J]. Neural Networks, 2021, 144: 247-259.

[5]JAIN S, SETH G, PARUTHI A, et al. Synthetic data augmentation for surface defect detection and classification using deep learning[J]. Journal of Intelligent Manufacturing, 2020，33:1007-1020.

[6]ZHANG H, PAN D, LIU J, et al. A novel MAS-GAN-based data synthesis method for object surface defectdetection[J]. Neurocomputing, 2022, 499:106-114.

[7]KIM J, JEONG K, CHOI H, et al. GAN-based anomaly detection in imbalance problems [C]//European Conference on Computer Vision, 2020:128-145.

[8]YE Z X, LIU M Q, ZHANG S L, et al. Dual-path GAN: a method for enhancing small-scale defect detection on metal images[C]//2022 41st Chinese Control Conference (CCC).[S.l.]: IEEE, 2022:6292-6297.

[9]RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks[J]. arXiv:1511.06434, 2015.

[10]陆顺成,姜小峰,石奇.基于改进DCGAN的汽车冷凝器图像生成方法[J].微电子学与计算机,2022,39(5):71-77.

LU C C, JIANG X F, SHI Q. Image generation method of automobile condenser based on improved DCGAN[J]. Microelectronics & Computer, 2022, 39(5): 71-77. (in Chinese)

[11]XIAO Z G, LU J, WANG X K, et al. WCE-DCGAN: a data augmentation method based on wireless capsule endoscopy images for gastrointestinal disease detection[J]. IET Image Processing, 2023,17(4):969-1306.

[12]GAO W, JIN H, YANG G J. Series arc fault diagnosis method of photovoltaic arrays based on GASF and improved DCGAN[J]. Advanced Engineering Informatics, 2022, 54:101809.

[13]RONNEBERGER O, FISCHER P, BROX T. U-net:convolutional networks for biomedical image segmentation[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention, 2015:234-241.

[14]CANNY J. A computational approach to edge detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1986, 8(6):679-698.

[15]MIYATO T, KATAOKA T, KOYAMA M, et al. Spectral normalization for generative adversarial networks[J]. arXiv:1802.05957, 2018.

[16]梁晨,王利斌,李卓群,等.生成对抗网络技术与研究进展[J].信息安全研究, 2022, 8(3):235-240.

LIANG C, WANG L B, LI Z Q,et al. The technology and research progress of generative adversarial networks[J]. Journal of Information Security Research, 2022,8(3):235-240. (in Chinese)

[17]MANGALAM K, GARG R. Overcoming mode collapse with adaptive multi adversarial training[J]. arXiv:2112.14406, 2021.

[18]米春风,卢琨,汪文艳,等.基于机器视觉的带钢表面缺陷检测研究进展[J].安徽工业大学学报(自然科学版), 2022, 39(2):180-188.

MI C F, LU K, WANG W Y,et al. Research progress of strip steel surface defect detection based on machine vision[J]. Journal of Anhui University of Technology(Natural Science), 2022, 39(2):180-188. (in Chinese)

[19]GAO X, DENG F, YUE X. Data augmentation in fault diagnosis based on the Wasserstein generative adversarial network with gradient penalty[J]. Neurocomputing, 2020, 396: 487-494.

[20]CHEN X, DUAN Y, HOUTHOOFT R, et al. InfoGAN: interpretable representation learning by information maximizing generative adversarial nets[J]. arXiv:1606.03657v1, 2016.

[21]HEUSEL M, RAMSAUER H, UNTERTHINER T, et al. GANs trained by a two time-scale update rule converge to a local nash equilibrium[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems, 2017:6629-6640.

[22]SHEIKH H R, BOVIK A C. Image information and visual quality[J]. IEEE Transactions on Image Processing, 2006, 15(2):430-444

[23]SHMELKOV K, CORDELIA S, ARTEEK A. How goodismy GAN? [J]. ECCV-European Conference on Computer Vision, 2018, 6:1-20.

[24]张成,潘明阳,高翊然,等.融合暗通道先验与循环生成对抗网络的航海图像去雾模型[J].大连海事大学学报, 2022, 48(4):84-93.

ZHANG C, PAN M Y, GAO Y R, et al. Maritime image dehazing model based on cycle generative adversarial network incorporating with dark channel prior[J]. Journal of Dalian Maritime University, 2022, 48(4):84-93. (in Chinese)

[25]JOCHER G, STOKEN A, BOROVEC J, et al.Ultralytics/yolov5: v5. 0-YOLOv5-P6 1280 models AWS Supervise. ly and YouTube integrations[J]. Zenodo, 2021: 11.doi: https://doi. org/10.5281/zenodo, 4679653.

[26]LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shotmultibox detector[J].Computer Vision-ECCV 2016, 2016: 21-37.

[27]GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015:1440-1448.

[28]CAI Z, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018:6154-6162.

[29]REDMON J, FARHADI A. Yolov3: an incremental improvement[J]. arXiv:1804.02767, 2018.

[30]ZHANG S, CHI C, YAO Y, et al. Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020:9759-9768.

[31]DUAN K, BAI S, XIE L, et al.Centernet: keypoint triplets for object detection[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019:6569-6578.

[32]KONG T, SUN F, LIU H, et al.Foveabox: beyound anchor-based object detection[J]. IEEE Transactions on Image Processing, 2020, 29:7389-7398.

Options

Outlines

模态框（Modal）标题

Abstract

Cite this article

References