Shukai Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Shukai Yang has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
- Xiaoqian Zhang
Southwest University of Science and Technology
5 shared publications - Yufeng Chen
School of Information and Control Engineering, Southwest University of Science and Technology, Mianyang, China
3 shared publications Feng Sun
Mianyang Central Hospital, School of Medicine, University of Electronic Science and Technology of China, Mianyang 621000, China; NHC Key Laboratory of Nuclear Technology Medical Transformation, Mianyang Central Hospital, Mianyang 621000, China
2 shared publications- Ying Zhou
Mianyang Central Hospital, School of Medicine, University of Electronic Science and Technology of China, Mianyang, China
2 shared publications - Jie Zhang
School of Information Engineering, Southwest University of Science and Technology, Mianyang 621010, People’s Republic of China
1 shared publication - Lei Zhang
School of Information Engineering, Southwest University of Science and Technology, Mianyang 621010, People’s Republic of China
1 shared publication
Latest Publications
- 22 April 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (9)
AM-DETR: Adaptive Multi-granularity Detection Transformer for PCB Defect Detection
To address the challenges of complex background textures, diverse defect shapes, scale variations, and low contrast between defects and background in printed circuit board (PCB) images, this paper proposes an Adaptive Multi-Granularity Detection Transformer (AM-DETR), specifically optimized for PCB defect detection. Firstly, to overcome the limitations of existing multi-scale fusion methods that often lack directional sensitivity and use static feature weighting, we introduce Multi-scale Salient Feature Fusion (MSFF), incorporating a direction-aware mechanism and dynamic weight allocation to significantly enhance focus on critical defect regions. Secondly, since conventional texture analysis techniques struggle to distinguish subtle surface defects from background noise, we propose the Surface Defect-aware Module (SDM), which improves perception of fine defects and suppresses noise through texture enhancement and adaptive feature modulation. Finally, to tackle inaccurate localization of defects with complex morphologies and varying scales, we design Multi-granularity Attention Fusion (MGAF), achieving precise defect localization and multi-level feature integration via collaborative global and local attention mechanisms. Experimental results demonstrate that AM-DETR achieves mAP50 scores of 99%, 96.7%, and 65.6% on the DeePPCB, PCB-DATASET, and pcb-aol datasets, respectively. The proposed method outperforms existing state-of-the-art approaches across all key performance metrics, achieving breakthroughs in detection accuracy and exhibiting exceptional adaptability in identifying defects of diverse sizes and complex morphologies.
- 22 April 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (9)
AM-DETR: Adaptive Multi-granularity Detection Transformer for PCB Defect Detection
To address the challenges of complex background textures, diverse defect shapes, scale variations, and low contrast between defects and background in printed circuit board (PCB) images, this paper proposes an Adaptive Multi-Granularity Detection Transformer (AM-DETR), specifically optimized for PCB defect detection. Firstly, to overcome the limitations of existing multi-scale fusion methods that often lack directional sensitivity and use static feature weighting, we introduce Multi-scale Salient Feature Fusion (MSFF), incorporating a direction-aware mechanism and dynamic weight allocation to significantly enhance focus on critical defect regions. Secondly, since conventional texture analysis techniques struggle to distinguish subtle surface defects from background noise, we propose the Surface Defect-aware Module (SDM), which improves perception of fine defects and suppresses noise through texture enhancement and adaptive feature modulation. Finally, to tackle inaccurate localization of defects with complex morphologies and varying scales, we design Multi-granularity Attention Fusion (MGAF), achieving precise defect localization and multi-level feature integration via collaborative global and local attention mechanisms. Experimental results demonstrate that AM-DETR achieves mAP50 scores of 99%, 96.7%, and 65.6% on the DeePPCB, PCB-DATASET, and pcb-aol datasets, respectively. The proposed method outperforms existing state-of-the-art approaches across all key performance metrics, achieving breakthroughs in detection accuracy and exhibiting exceptional adaptability in identifying defects of diverse sizes and complex morphologies.
- 23 October 2025
- Published by Elsevier in Neural Networks
- Vol. 195, 108230
DNUNet: A lightweight adaptive medical image segmentation network based on dual-path multilevel interactive convolution and norm sparse fusion module
- 23 October 2025
- Published by Elsevier in Neural Networks
- Vol. 195, 108230
DNUNet: A lightweight adaptive medical image segmentation network based on dual-path multilevel interactive convolution and norm sparse fusion module
- 7 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (1) , 521-534
TBE-Net: A Deep Network Based on Tree-Like Branch Encoder for Medical Image Segmentation
In recent years, encoder-decoder-based network structures have been widely used in designing medical image segmentation models. However, these methods still face some limitations: 1) The network's feature extraction capability is limited, primarily due to insufficient attention to the encoder, resulting in a failure to extract rich and effective features. 2) Unidirectional stepwise decoding of smaller-sized feature maps restricts segmentation performance. To address the above limitations, we propose an innovative Tree-like Branch Encoder Network (TBE-Net), which adopts a tree-like branch encoder to better perform feature extraction and preserve feature information. Additionally, we introduce the Depth and Width Expansion (D-WE) module to expand the network depth and width at low parameter cost, thereby enhancing network performance. Furthermore, we design a Deep Aggregation Module (DAM) to better aggregate and process encoder features. Subsequently, we directly decode the aggregated features to generate the segmentation map. The experimental results show that, compared to other advanced algorithms, our method, with the lowest parameter cost, achieved improvements in the IoU metric on the TNBC, PH2, CHASE-DB1, STARE, and COVID-19-CT-Seg datasets by 1.6%, 0.46%, 0.81%, 1.96%, and 0.86%, respectively
- 7 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (1) , 521-534
TBE-Net: A Deep Network Based on Tree-Like Branch Encoder for Medical Image Segmentation
In recent years, encoder-decoder-based network structures have been widely used in designing medical image segmentation models. However, these methods still face some limitations: 1) The network's feature extraction capability is limited, primarily due to insufficient attention to the encoder, resulting in a failure to extract rich and effective features. 2) Unidirectional stepwise decoding of smaller-sized feature maps restricts segmentation performance. To address the above limitations, we propose an innovative Tree-like Branch Encoder Network (TBE-Net), which adopts a tree-like branch encoder to better perform feature extraction and preserve feature information. Additionally, we introduce the Depth and Width Expansion (D-WE) module to expand the network depth and width at low parameter cost, thereby enhancing network performance. Furthermore, we design a Deep Aggregation Module (DAM) to better aggregate and process encoder features. Subsequently, we directly decode the aggregated features to generate the segmentation map. The experimental results show that, compared to other advanced algorithms, our method, with the lowest parameter cost, achieved improvements in the IoU metric on the TNBC, PH2, CHASE-DB1, STARE, and COVID-19-CT-Seg datasets by 1.6%, 0.46%, 0.81%, 1.96%, and 0.86%, respectively
- 1 August 2024
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 180, 108947
FCSU-Net: A novel full-scale Cross-dimension Self-attention U-Net with collaborative fusion of multi-scale feature for medical image segmentation
- 1 August 2024
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 180, 108947
FCSU-Net: A novel full-scale Cross-dimension Self-attention U-Net with collaborative fusion of multi-scale feature for medical image segmentation
- 15 January 2024
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 170, 108009
FAFS-UNet: Redesigning skip connections in UNet with feature aggregation and feature selection
- 15 January 2024
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 170, 108009
FAFS-UNet: Redesigning skip connections in UNet with feature aggregation and feature selection