Wenzhu Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Wenzhu Yang has a total of 42 co-authors. Here are the top co-authors by number of shared publications:
Yanyan Jiao
School of Cyber Security and Computer, Hebei University, Baoding, China
5 shared publicationsLei Geng
School of Cyber Security and Computer, Hebei University, Baoding, China
4 shared publicationsShuang Zeng
School of Cyber Security and Computer, Hebei University, Baoding, China
4 shared publications- Ruru Gou
School of Cyber Security and Computer, Hebei University, Baoding 071002, China
3 shared publications Xiaoyang Li
School of Cyber Security and Computer Hebei University 071000 Baoding Hebei China
3 shared publications- Yunfeng Yuan
School of Cyber Security and Computer, Hebei University, Baoding 071002, China
3 shared publications
Latest Publications
- 19 July 2026
- Published by Springer Nature
- p. 125-138
An Aerial Image Object Detection Method Based on Fine-Grained Feature Extraction and Feature Aware
- 19 July 2026
- Published by Springer Nature
- p. 125-138
An Aerial Image Object Detection Method Based on Fine-Grained Feature Extraction and Feature Aware
- Open Access
- 13 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 201-204
A Genetic Programming Multi-Agent Topology Optimization Method under Stability Constraints
- 13 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 201-204
Open AccessA Genetic Programming Multi-Agent Topology Optimization Method under Stability Constraints
- Open Access
Context-Aware Network Based on Multi-scale Spatio-temporal Attention for Action Recognition in Videos
Action recognition is a critical task in video understanding, requiring the comprehensive capture of spatio-temporal cues across various scales. However, existing methods often overlook the multi-granularity nature of actions. To address this limitation, we introduce the Context-Aware Network (CAN). CAN consists of two core modules: the Multi-scale Temporal Cue Module (MTCM) and the Group Spatial Cue Module (GSCM). MTCM effectively extracts temporal cues at multiple scales, capturing both fast-changing motion details and overall action flow. GSCM, on the other hand, extracts spatial cues at different scales by grouping feature maps and applying specialized extraction methods to each group. Experiments conducted on five benchmark datasets (Something-Something V1 and V2, Diving48, Kinetics-400, and UCF101) demonstrate the effectiveness of CAN. Our approach achieves competitive performance, outperforming most mainstream methods, with accuracies of 50.4% on Something-Something V1, 63.9% on Something-Something V2, 88.4% on Diving48, 74.9% on Kinetics-400, and 86.9% on UCF101. These results highlight the importance of capturing multi-scale spatio-temporal cues for robust action recognition.
Open AccessContext-Aware Network Based on Multi-scale Spatio-temporal Attention for Action Recognition in Videos
Action recognition is a critical task in video understanding, requiring the comprehensive capture of spatio-temporal cues across various scales. However, existing methods often overlook the multi-granularity nature of actions. To address this limitation, we introduce the Context-Aware Network (CAN). CAN consists of two core modules: the Multi-scale Temporal Cue Module (MTCM) and the Group Spatial Cue Module (GSCM). MTCM effectively extracts temporal cues at multiple scales, capturing both fast-changing motion details and overall action flow. GSCM, on the other hand, extracts spatial cues at different scales by grouping feature maps and applying specialized extraction methods to each group. Experiments conducted on five benchmark datasets (Something-Something V1 and V2, Diving48, Kinetics-400, and UCF101) demonstrate the effectiveness of CAN. Our approach achieves competitive performance, outperforming most mainstream methods, with accuracies of 50.4% on Something-Something V1, 63.9% on Something-Something V2, 88.4% on Diving48, 74.9% on Kinetics-400, and 86.9% on UCF101. These results highlight the importance of capturing multi-scale spatio-temporal cues for robust action recognition.
- 26 November 2025
- Published by Springer Nature in Multimedia Systems
- Vol. 31 (6) , 470
Unidirectional guidance network for enhanced small object detection in UAV imagery
- 26 November 2025
- Published by Springer Nature in Multimedia Systems
- Vol. 31 (6) , 470
Unidirectional guidance network for enhanced small object detection in UAV imagery
- 23 June 2025
- Published by Elsevier in Knowledge-Based Systems
- Vol. 326
CLIP-MEI: Exploit more effective information for few-shot action recognition
- 23 June 2025
- Published by Elsevier in Knowledge-Based Systems
- Vol. 326
CLIP-MEI: Exploit more effective information for few-shot action recognition
- Open Access
- 16 April 2025
- Published by MDPI AG in Information
- Vol. 16 (4) , 315
Camera Pose Generation Based on Unity3D
Deep learning models performing complex tasks require the support of datasets. With the advancement of virtual reality technology, the use of virtual datasets in deep learning models is becoming more and more widespread. Indoor scenes represents a significant area of interest for the application of machine vision technologies. Existing virtual indoor datasets exhibit deficiencies with regard to camera poses, resulting in problems such as occlusion, object omission, and objects having too small of a proportion of the image, and perform poorly in the training for object detection and simultaneous localization and mapping (SLAM) tasks. Aiming at the problems regarding the capacity of cameras to comprehensively capture scene objects, this study presents an enhanced algorithm based on rapidly exploring random tree star (RRT*) for the generation of camera poses in a 3D indoor scene. Meanwhile, in order to generate multimodal data for various deep learning tasks, this study designs an automatic image acquisition module under the Unity3D platform. The experimental results from running the model on several mainstream virtual indoor datasets—such as 3D-FRONT and Hypersim—indicate that the image sequences generated in this study show enhancements in terms of object capture rate and efficiency. Even in cluttered environments such as those in SceneNet RGB-D, the object capture rate remains stable at around 75%. Compared with the image sequences from the original datasets, those generated in this study achieve improvements in the object detection and SLAM tasks, with increases of up to approximately 30% in mAP for the YOLOv10 object detection task and up to approximately 10% in SR for the ORB-SLAM algorithm.
- 16 April 2025
- Published by MDPI AG in Information
- Vol. 16 (4) , 315
Open AccessCamera Pose Generation Based on Unity3D
Deep learning models performing complex tasks require the support of datasets. With the advancement of virtual reality technology, the use of virtual datasets in deep learning models is becoming more and more widespread. Indoor scenes represents a significant area of interest for the application of machine vision technologies. Existing virtual indoor datasets exhibit deficiencies with regard to camera poses, resulting in problems such as occlusion, object omission, and objects having too small of a proportion of the image, and perform poorly in the training for object detection and simultaneous localization and mapping (SLAM) tasks. Aiming at the problems regarding the capacity of cameras to comprehensively capture scene objects, this study presents an enhanced algorithm based on rapidly exploring random tree star (RRT*) for the generation of camera poses in a 3D indoor scene. Meanwhile, in order to generate multimodal data for various deep learning tasks, this study designs an automatic image acquisition module under the Unity3D platform. The experimental results from running the model on several mainstream virtual indoor datasets—such as 3D-FRONT and Hypersim—indicate that the image sequences generated in this study show enhancements in terms of object capture rate and efficiency. Even in cluttered environments such as those in SceneNet RGB-D, the object capture rate remains stable at around 75%. Compared with the image sequences from the original datasets, those generated in this study achieve improvements in the object detection and SLAM tasks, with increases of up to approximately 30% in mAP for the YOLOv10 object detection task and up to approximately 10% in SR for the ORB-SLAM algorithm.
- 27 February 2025
- Published by Springer Nature
- p. 174-185
LSBNet: Lightweight Symmetrically Balanced Network for Real-Time Semantic Segmentation
- 27 February 2025
- Published by Springer Nature
- p. 174-185
LSBNet: Lightweight Symmetrically Balanced Network for Real-Time Semantic Segmentation
- Open Access
- 6 February 2025
- Published by Springer Nature in Complex & Intelligent Systems
- Vol. 11 (3) , 1-18
Manet: motion-aware network for video action recognition
- 6 February 2025
- Published by Springer Nature in Complex & Intelligent Systems
- Vol. 11 (3) , 1-18
Open AccessManet: motion-aware network for video action recognition
- 12 November 2024
- Published by Springer Nature
- p. 195-207
Action Recognition Based on Multi-perspective Feature Excitation
- 12 November 2024
- Published by Springer Nature
- p. 195-207
Action Recognition Based on Multi-perspective Feature Excitation
- 1 November 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Electronic Imaging
- Vol. 33 (06) , 063011-22
CDANet: a small object detection model for unmanned aerial vehicle images based on cross-layer guidance and skip dilated fusion
Detecting small objects in unmanned aerial vehicle (UAV) pictures is a vital but demanding task in the field of computer vision due to factors such as object size, image degradation, and computational resource limitations. Most existing algorithms improve object detection accuracy by improving feature fusion structures and incorporating single-layer attention modules. However, the weight of single-layer attention is computed only based on the original input, which leads to insufficient exploration of correlations among cross-layer features. To remedy the above issues, we design a novel multi-scale UAV image object detection model called CDANet. First, to fully utilize the scattered detail features in the shallow layers, a cross-layer attention module named the enhanced guided attention module is designed. Second, a simplified depth-wise separable dilated spatial pyramid pooling-fast module is built to overcome the shortcomings of max-pooling operation during feature fusion. Third, an efficient coordinate guidance module is designed as the basic module for guiding the inter-channel information interaction. Fourth, a cross-spatial learning encoder module is designed to encode cross-channel information with an efficient multi-scale attention module. Last, to optimize the learning process by dynamically regulating the attention toward diverse objects, an advanced loss function Wise-IoU v3 is introduced. The results show that, on the VisDrone2021-DET dataset, the average precision and mean average precision of the proposed method are superior to YOLOv8, which are respectively improved by 3.9% and 4.9%.
- 1 November 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Electronic Imaging
- Vol. 33 (06) , 063011-22
CDANet: a small object detection model for unmanned aerial vehicle images based on cross-layer guidance and skip dilated fusion
Detecting small objects in unmanned aerial vehicle (UAV) pictures is a vital but demanding task in the field of computer vision due to factors such as object size, image degradation, and computational resource limitations. Most existing algorithms improve object detection accuracy by improving feature fusion structures and incorporating single-layer attention modules. However, the weight of single-layer attention is computed only based on the original input, which leads to insufficient exploration of correlations among cross-layer features. To remedy the above issues, we design a novel multi-scale UAV image object detection model called CDANet. First, to fully utilize the scattered detail features in the shallow layers, a cross-layer attention module named the enhanced guided attention module is designed. Second, a simplified depth-wise separable dilated spatial pyramid pooling-fast module is built to overcome the shortcomings of max-pooling operation during feature fusion. Third, an efficient coordinate guidance module is designed as the basic module for guiding the inter-channel information interaction. Fourth, a cross-spatial learning encoder module is designed to encode cross-channel information with an efficient multi-scale attention module. Last, to optimize the learning process by dynamically regulating the attention toward diverse objects, an advanced loss function Wise-IoU v3 is introduced. The results show that, on the VisDrone2021-DET dataset, the average precision and mean average precision of the proposed method are superior to YOLOv8, which are respectively improved by 3.9% and 4.9%.