Research on a Lightweight PCB Detection Algorithm Based on AE-YOLO DOI Creative Commons
Yuanyuan Wang, Y. G. Li,

Dipu Md Sharid Kayes

и другие.

IEEE Access, Год журнала: 2024, Номер 12, С. 109367 - 109379

Опубликована: Янв. 1, 2024

The attention enhancement YOLO printed circuit board (PCB) defect detection algorithm AE-YOLO, which improves YOLOv8, is proposed to improve the current slow speed of PCB problems, such as high missed or false rates and low accuracy. First, in backbone network, CoT Net used instead original feature extraction network reduce number parameters model its while maintaining accuracy much possible. Then, SPPFS module last layer enhance model's ability extract global information, fuse features, use rich primary semantic information pave way for subsequent classification positioning. Finally, CC3 perceive high-level help decoupled head better perform target prediction positioning, comprehensiveness model, provide with continuous performance improvements. Compared YOLOv8 AE-YOLO compresses by 16%, increases 2.9%, recall rate 3.3%. This provides a more efficient method detection.

Язык: Английский

YOLO-HMC: An Improved Method for PCB Surface Defect Detection DOI
Minghao Yuan, Yongbing Zhou, Xiaoyu Ren

и другие.

IEEE Transactions on Instrumentation and Measurement, Год журнала: 2024, Номер 73, С. 1 - 11

Опубликована: Янв. 1, 2024

The surface defects of printed circuit boards (PCBs) generated during the manufacturing process have an adverse effect on product quality, which further directly affects stability and reliability equipment performance. However, there are still great challenges in accurately recognizing tiny PCB under complex background due to its compact layout. To address problem, a novel YOLO-HorNet-MCBAM-CARAFE (YOLO-HMC) network based improved YOLOv5 framework is proposed this article identify tiny-size defect more efficiently with fewer model parameters. First, backbone part adopts HorNet for enhancing feature extraction ability deepening information interaction. Second, multiple convolutional block attention module (MCBAM) designed improve highlight location from highly similar substrate background. Third, content-aware reassembly features (CARAFE) used replace up-sampling layer fully aggregating contextual semantic images large receptive field. Moreover, aiming at difference between detection natural detection, original head (DH) optimized ensure that can detect defects. Extensive experiments public datasets demonstrated significant advantage compared several state-of-the-art models, whose mean average precision (mAP) reach 98.6%, verifying accuracy applicability YOLO-HMC.

Язык: Английский

Процитировано

33
RanMerFormer: Randomized vision transformer with token merging for brain tumor classification DOI Creative Commons
Jian Wang, Siyuan Lu, Shuihua Wang‎

и другие.

Neurocomputing, Год журнала: 2024, Номер 573, С. 127216 - 127216

Опубликована: Янв. 5, 2024

Brains are the control center of nervous system in human bodies, and brain tumor is one most deadly diseases. Currently, magnetic resonance imaging (MRI) effective way to tumors early detection clinical diagnoses due its superior quality for soft tissues. Manual analysis MRI error-prone which depends on empirical experience fatigue state radiologists a large extent. Computer-aided diagnosis (CAD) systems becoming more impactful because they can provide accurate prediction results based medical images with advanced techniques from computer vision. Therefore, novel CAD method classification named RanMerFormer presented this paper. A pre-trained vision transformer used as backbone model. Then, merging mechanism proposed remove redundant tokens transformer, improves computing efficiency substantially. Finally, randomized vector functional-link serves head RanMerFormer, be trained swiftly. All simulation obtained two public benchmark datasets, reveal that achieve state-of-the-art performance classification. The applied real-world scenarios assist diagnosis.

Язык: Английский

Процитировано

30
Development and challenges of object detection: A survey DOI
Zonghui Li, Yongsheng Dong,

Longchao Shen

и другие.

Neurocomputing, Год журнала: 2024, Номер 598, С. 128102 - 128102

Опубликована: Июнь 22, 2024

Язык: Английский

Процитировано

8
LiFSO-Net: A lightweight feature screening optimization network for complex-scale flat metal defect detection DOI
Hao Zhong, Ling Xiao, Haifeng Wang

и другие.

Knowledge-Based Systems, Год журнала: 2024, Номер unknown, С. 112520 - 112520

Опубликована: Сен. 1, 2024

Язык: Английский

Процитировано

7
DHNet: a surface defect detection model utilizing multi-scale convolutional kernels DOI
Yingying Zhang, Shuo Wang, Jinhai Wang

и другие.

Journal of Real-Time Image Processing, Год журнала: 2025, Номер 22(1)

Опубликована: Янв. 17, 2025

Язык: Английский

Процитировано

1
Joint learning of multi-level dynamic brain networks for autism spectrum disorder diagnosis DOI Creative Commons
Na Li, Jinjie Xiao, Ning Mao

и другие.

Computers in Biology and Medicine, Год журнала: 2024, Номер 171, С. 108054 - 108054

Опубликована: Фев. 8, 2024

Graph convolutional networks (GCNs), with their powerful ability to model non-Euclidean graph data, have shown advantages in learning representations of brain networks. However, considering the complexity, multilayeredness, and spatio-temporal dynamics activities, we identified two limitations current GCN-based research on networks: 1) Most studies focused unidirectional information transmission across network levels, neglecting joint or bidirectional exchange among 2) existing models determine node neighborhoods by thresholding simply binarizing network, which leads loss edge weight weakens model's sensitivity important network. To address above issues, propose a multi-level dynamic architecture based GCN for autism spectrum disorder (ASD) diagnosis. Specifically, firstly, constructing different-level Then, utilizing interactive these Finally, designing an self-attention mechanism assign different weights inter-node connections, allows us pick out crucial features ASD Our proposed method achieves accuracy 81.5 %. The results demonstrate that our enables transfer high-order low-order information, facilitating complementarity between levels Additionally, use enhances representation capability ASD-related features.

Язык: Английский

Процитировано

5
A real-time PCB defect detection model based on enhanced semantic information fusion DOI
Tangyu Ji, Qian Zhao, Kang An

и другие.

Signal Image and Video Processing, Год журнала: 2024, Номер 18(6-7), С. 4945 - 4959

Опубликована: Июнь 22, 2024

Язык: Английский

Процитировано

4
ACAT-transformer: Adaptive classifier with attention-wise transformation for few-sample surface defect recognition DOI
Zhaofu Li, Liang Gao, Xinyu Li

и другие.

Advanced Engineering Informatics, Год журнала: 2024, Номер 61, С. 102527 - 102527

Опубликована: Апрель 17, 2024

Язык: Английский

Процитировано

3
REDef-DETR: real-time and efficient DETR for industrial surface defect detection DOI

Dejian Li,

Changhong Jiang,

Tielin Liang

и другие.

Measurement Science and Technology, Год журнала: 2024, Номер 35(10), С. 105411 - 105411

Опубликована: Июль 9, 2024

Abstract Industrial surface defect detection is an important part of industrial production, which aims to identify and detecting various defects on the product ensure quality meet customer requirements. With development deep learning image processing technologies, methods based computer vision has become mainstream method. However, prevalent convolutional neural network-based also have many problems. For example, these rely post-processing Non-Maximum Suppression poor ability for small targets, affects speed accuracy in scenarios. Therefore, we propose a novel DEtection TRansformer-based Firstly, Multi-scale Contextual Information Dilated module fuse it into backbone. The mainly composed large kernel convolutions, expand receptive field model, thus reducing leakage rate model. Moreover, design efficient encoder contains two modules, namely feature enhancement cascaded group attention fusion content-aware. former effectively enhances high-level semantic information extracted by backbone, enabling model better interpret features, can improve problem high computational cost transformer encoder, increasing speed. latter performs multi-scale across scales, improving small-size defects. Experimental results show that proposed method achieves 80.6%mAP 80.3FPS NEU-DET, 98.0%mAP 79.4FPS PCB-DET. Our exhibits excellent performance real-time capability needs detection.

Язык: Английский

Процитировано

3
LSKA-YOLOv8: A lightweight steel surface defect detection algorithm based on YOLOv8 improvement DOI Creative Commons
Jun Tie,

Chengao Zhu,

Zheng Lu

и другие.

Alexandria Engineering Journal, Год журнала: 2024, Номер 109, С. 201 - 212

Опубликована: Сен. 4, 2024

Язык: Английский

Процитировано

3