Pixel-Wise Defect Detection by CNNs without Manually Labeled Training Data

Haselmann, Matthias; Gruber, Dieter P.

doi:10.1080/08839514.2019.1583862

Cited by 21 publications

(6 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In 2019, M. Haselmann and D.P. Gruber presented CNNs for defect detection on decorative plastic parts without manually labeled training data [2]. Madhuri Pathak et al developed a method for detecting rail foot defects based on the propagation of laser-induced ultrasonic waveguides as a tool to detect cracks in rail footings [3].…”

Section: Related Workmentioning

confidence: 99%

Defect detection of injection-molded parts based on improved-YOLOv5

Liang

Chen

Xie

et al. 2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Quality inspection with injection-molded parts (IMPs) is an essential part of the process of producing IMPs. To improve production quality, we propose an improved-YOLOv5 model to detect the object defection of the IMPs. Different from the original YOLOv5, we replace the coupled detection head (CH) with the independent decoupled head (IDCH), while adding the Squeeze-and-Excitation attention mechanism (SE) into the regression branch. Furthermore, a defective dataset of IMPs is constructed within the real-world production line, including 1357 IMPs images. The experimental results demonstrate the effectiveness of the improve-YOLOv5. Compared with YOLOv5, the precision and recall of the modified algorithm in this paper are increased by 2.5 % and 4.7 %, respectively. For difficult categories, precision and recall are increased by 8.7 % and 13.5 %, respectively.

show abstract

Section: Related Workmentioning

confidence: 99%

Defect detection of injection-molded parts based on improved-YOLOv5

Liang

Chen

Xie

et al. 2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

show abstract

“…Data augmentation technic can directly increase the number of defect samples. Haselmann et al 4 inject synthetized defects into fault-free surface images for training CNN. Unsupervised or semi-supervised learning algorithms can reduce the reliance on data.…”

Section: Introductionmentioning

confidence: 99%

A multi-source transfer learning-based weighted network for small sample defect inspection

Hong

Yan

et al. 2023

Optoelectronic Imaging and Multimedia Technology IX

View full text Add to dashboard Cite

Defect inspection is indispensable process in manufacturing, and automatic optical inspection (AOI) has been rapidly applied to various areas. In AOI, artificial intelligence (AI) based deep learning methods are more and more advantageous in many fields. However, obtainment of high-performance deep learning algorithms always requires a large amount of training data, while defect samples are often scarce. So small sample has become one of the key problems in the industrial application of deep learning algorithms. Transfer learning enable us to utilize the knowledge of source domains to improve performance on target domain, which could be used to tackle the small sample problem intuitively. Therefore, this paper proposes a defect inspection network which is based on one of the transfer learning techniques: domain adaptation. We name the network as multi-source and multi-scale weighted domain adaptation network which is based on adversarial learning. Firstly, three adversarial domain adaptation modules are proposed to align feature distributions between multi-source domains and target domain under three scales, which make the backbone extract domain-invariant features. Simultaneously, the weights of domain adaptation module under each scale are set reasonably. Secondly, in order to reduce the effect of negative transfer, a novel similarity weight is proposed, which is applied on domain adaptation modules. Finally, experiments are carried out to prove the effectiveness of our method. The results show that our method can improve the mean average precision(mAP) from 62.3 to 78.5 in the case of 40 samples available for 4 defect categories, which surpasses other counterparts.

show abstract

“…Anomaly detection methods focuses on the detection of previously unseen samples. [13] has been used GANs to generate synthetic defects similar to the patches of the test set as possible. Similarly, the hologram verification task is an anomaly detection problem where fake holograms are usually cannot distinguished by the naked eye.…”

Section: Introductionmentioning

confidence: 99%

Open-Set Learning-Based Hologram Verification System Using Generative Adversarial Networks

2022

IEEE Access

View full text Add to dashboard Cite

show abstract

Pixel-Wise Defect Detection by CNNs without Manually Labeled Training Data

Cited by 21 publications

References 27 publications

Defect detection of injection-molded parts based on improved-YOLOv5

Defect detection of injection-molded parts based on improved-YOLOv5

A multi-source transfer learning-based weighted network for small sample defect inspection

Open-Set Learning-Based Hologram Verification System Using Generative Adversarial Networks

Contact Info

Product

Resources

About