Guifu Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Guifu Zhu has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Hairui Wang
Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650504, China
8 shared publications Hairui Wang
Kunming University of Science and Technology
4 shared publications- Wei Dai
Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, China
3 shared publications - Xiaoli Zhang
Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming, China
3 shared publications - Ya Li
Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650504, China
3 shared publications - Jialei Nie
School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650500, China
2 shared publications
Latest Publications
- Open Access
- 17 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (18) , 9231
Lightweight ACG-YOLO with Collaborative Attention Mechanisms for Real-Time Public Safety Surveillance and Edge Deployment
Real-time detection of violent acts and hazardous weapons in edge-deployed public safety surveillance remains challenging because resource-constrained devices must balance detection accuracy and computational efficiency, particularly for small-sized weapon targets in complex multi-person interaction scenes. To address these issues, this paper proposes ACG-YOLO, a lightweight collaborative detection framework built upon the YOLOv11n baseline. The model adopts a three-level design involving efficient feature downsampling, spatial localization enhancement, and global context modeling. Specifically, an ADown module reduces information loss during downsampling through a dual-branch structure, helping preserve fine-grained features of small targets; a Coordinate Attention mechanism enhances spatial localization by encoding horizontal and vertical positional information; and a Global Context Enhancement Module (GCEM) incorporates lightweight global contextual information to improve feature representation in complex violent interaction scenes. Experimental results on a violence and weapon detection dataset demonstrate that ACG-YOLO achieves an mAP@0.5 of 89.2%, representing an improvement of 2.9 percentage points over the baseline YOLOv11n, while reducing the parameter count by 19.8% and computational cost by 19.0%, with only 2.07 M parameters and 5.1 GFLOPs. Ablation experiments further examine the contributions of the individual components and their combinations. This research provides a practical solution for real-time early warning and emergency response in resource-constrained public safety surveillance scenarios.
- 17 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (18) , 9231
Open AccessLightweight ACG-YOLO with Collaborative Attention Mechanisms for Real-Time Public Safety Surveillance and Edge Deployment
Real-time detection of violent acts and hazardous weapons in edge-deployed public safety surveillance remains challenging because resource-constrained devices must balance detection accuracy and computational efficiency, particularly for small-sized weapon targets in complex multi-person interaction scenes. To address these issues, this paper proposes ACG-YOLO, a lightweight collaborative detection framework built upon the YOLOv11n baseline. The model adopts a three-level design involving efficient feature downsampling, spatial localization enhancement, and global context modeling. Specifically, an ADown module reduces information loss during downsampling through a dual-branch structure, helping preserve fine-grained features of small targets; a Coordinate Attention mechanism enhances spatial localization by encoding horizontal and vertical positional information; and a Global Context Enhancement Module (GCEM) incorporates lightweight global contextual information to improve feature representation in complex violent interaction scenes. Experimental results on a violence and weapon detection dataset demonstrate that ACG-YOLO achieves an mAP@0.5 of 89.2%, representing an improvement of 2.9 percentage points over the baseline YOLOv11n, while reducing the parameter count by 19.8% and computational cost by 19.0%, with only 2.07 M parameters and 5.1 GFLOPs. Ablation experiments further examine the contributions of the individual components and their combinations. This research provides a practical solution for real-time early warning and emergency response in resource-constrained public safety surveillance scenarios.
- Open Access
- 15 September 2026
- Published by MDPI AG in Electronics
- Vol. 15 (18) , 4198
BOM-YOLOv12n: A Lightweight Detection Model for Protective Equipment in Smart Construction Sites
In complex construction site environments, protective equipment detection suffers from large target scale variations, frequent occlusions, and strong background interference, making it difficult for existing lightweight models to balance detection accuracy with computational cost. To address these challenges, an enhanced lightweight detection model termed BOM-YOLOv12n is introduced. Based on YOLOv12n, a BN-calibrated re-parameterized convolution (BCRepConv) is first designed to reduce the feature distribution gap between training and inference and enhance small-target feature representation. A multi-dimensional attention module ODC2f is then introduced to strengthen local discriminative responses under occlusion. Finally, the MPDIoU loss based on minimum point distance is adopted to improve bounding box regression accuracy. Experimental results on the HardHat-Vest dataset show that BOM-YOLOv12n achieves an mAP0.5 of 87.3% and an mAP0.5:0.95 of 58.5%, with only 2.3 M parameters and 6.2 GFLOPs, achieving a favorable balance between detection accuracy and computational efficiency compared to the baseline and other mainstream lightweight detectors. The experimental results indicate that BOM-YOLOv12n can serve as a competitive and efficient solution for real-time protective equipment detection in smart construction sites.
- 15 September 2026
- Published by MDPI AG in Electronics
- Vol. 15 (18) , 4198
Open AccessBOM-YOLOv12n: A Lightweight Detection Model for Protective Equipment in Smart Construction Sites
In complex construction site environments, protective equipment detection suffers from large target scale variations, frequent occlusions, and strong background interference, making it difficult for existing lightweight models to balance detection accuracy with computational cost. To address these challenges, an enhanced lightweight detection model termed BOM-YOLOv12n is introduced. Based on YOLOv12n, a BN-calibrated re-parameterized convolution (BCRepConv) is first designed to reduce the feature distribution gap between training and inference and enhance small-target feature representation. A multi-dimensional attention module ODC2f is then introduced to strengthen local discriminative responses under occlusion. Finally, the MPDIoU loss based on minimum point distance is adopted to improve bounding box regression accuracy. Experimental results on the HardHat-Vest dataset show that BOM-YOLOv12n achieves an mAP0.5 of 87.3% and an mAP0.5:0.95 of 58.5%, with only 2.3 M parameters and 6.2 GFLOPs, achieving a favorable balance between detection accuracy and computational efficiency compared to the baseline and other mainstream lightweight detectors. The experimental results indicate that BOM-YOLOv12n can serve as a competitive and efficient solution for real-time protective equipment detection in smart construction sites.
- 21 February 2026
- Published by Taylor & Francis in Nondestructive Testing and Evaluation
- p. 1-31
TFMPINN: a meta-learning approach for remaining useful life prediction via time-domain latent encoding and frequency-domain physics-informed neural networks
1. Aero-engines, as the core components of aircraft propulsion systems, possess highly complex internal structures and operate under harsh environmental conditions. As operating time increases, int...
- 21 February 2026
- Published by Taylor & Francis in Nondestructive Testing and Evaluation
- p. 1-31
TFMPINN: a meta-learning approach for remaining useful life prediction via time-domain latent encoding and frequency-domain physics-informed neural networks
1. Aero-engines, as the core components of aircraft propulsion systems, possess highly complex internal structures and operate under harsh environmental conditions. As operating time increases, int...
- Open Access
- 4 December 2025
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Variance-stabilized cognitive diagnosis via GCN-enhanced graph attention with adaptive relation pruning
Cognitive diagnosis, pivotal for AI-enhanced learning, faces the challenge of fusing robust modeling of complex student-exercise-knowledge relationships with psychometric theoretical consistency: (1) performance degradation from over-smoothing induced by polarized knowledge associations in graph attention networks, and (2) limited interpretability due to the opaque nature of deep learning architectures. We present GEAR-CD, a GCN-enhanced graph attention framework with adaptive relation pruning that addresses these limitations through three key innovations: First, hierarchical graph attention unifies interaction modeling across heterogeneous relationships (knowledge concepts, exercises, and learners). Second, GCN-based convolutional kernels with automated edge pruning mitigate attention-driven over-smoothing. Third, theoretically-grounded design ensures alignment with Item Response Theory (IRT) principles. Comprehensive evaluations on Junyi, Assistments and EdNet datasets demonstrate GEAR-CD’s superiority, achieving 6.85% and 2.05% relative improvements over baseline and state-of-the-art models respectively in prediction accuracy (ACC=81.67%), while maintaining exceptional stability (\(\sigma<\)0.0001). Visualization analyses confirm both diagnostic robustness (via radar plots) and theoretical validity (t-SNE manifolds linearity \(\hbox {R}^{2}\)=0.93), establishing GEAR-CD as a credible solution for operational learning systems.- 4 December 2025
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessVariance-stabilized cognitive diagnosis via GCN-enhanced graph attention with adaptive relation pruning
Cognitive diagnosis, pivotal for AI-enhanced learning, faces the challenge of fusing robust modeling of complex student-exercise-knowledge relationships with psychometric theoretical consistency: (1) performance degradation from over-smoothing induced by polarized knowledge associations in graph attention networks, and (2) limited interpretability due to the opaque nature of deep learning architectures. We present GEAR-CD, a GCN-enhanced graph attention framework with adaptive relation pruning that addresses these limitations through three key innovations: First, hierarchical graph attention unifies interaction modeling across heterogeneous relationships (knowledge concepts, exercises, and learners). Second, GCN-based convolutional kernels with automated edge pruning mitigate attention-driven over-smoothing. Third, theoretically-grounded design ensures alignment with Item Response Theory (IRT) principles. Comprehensive evaluations on Junyi, Assistments and EdNet datasets demonstrate GEAR-CD’s superiority, achieving 6.85% and 2.05% relative improvements over baseline and state-of-the-art models respectively in prediction accuracy (ACC=81.67%), while maintaining exceptional stability (\(\sigma<\)0.0001). Visualization analyses confirm both diagnostic robustness (via radar plots) and theoretical validity (t-SNE manifolds linearity \(\hbox {R}^{2}\)=0.93), establishing GEAR-CD as a credible solution for operational learning systems. - Open Access
- 15 October 2025
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 37 (8) , 1-18
DA-TCKT: Adaptive time convolution knowledge tracing method based on discrete cosine transform attention mechanism
- 15 October 2025
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 37 (8) , 1-18
Open AccessDA-TCKT: Adaptive time convolution knowledge tracing method based on discrete cosine transform attention mechanism
- Open Access
- 2 September 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (17) , 7723
Method for Remaining Useful Life Prediction of Turbofan Engines Combining Adam Optimization-Based Self-Attention Mechanism with Temporal Convolutional Networks
Conducting the remaining useful life (RUL) prediction for an aircraft engines is of significant importance in enhancing aircraft operation safety and formulating reasonable maintenance plans. Addressing the issue of low prediction model accuracy due to traditional neural networks’ inability to fully extract key features, this paper proposes an engine RUL prediction model based on the adaptive moment estimation (Adam) optimized self-attention mechanism–temporal convolutional network (SAM-TCN) neural network. Firstly, the raw data monitored by sensors are normalized, and RUL labels are set. A sliding window is utilized for overlapping sampling of the data, capturing more temporal features while eliminating data dimensionality. Secondly, the SAM-TCN neural network prediction model is constructed. The temporal convolutional network (TCN) neural network is used to capture the temporal dependency between data, solving the mapping relationship of engine degradation characteristics. A self-attention mechanism (SAM) is employed to adaptively assign different weight contributions to different input features. In the experiments, the root mean square error (RMSE) values on four datasets are 11.50, 16.45, 11.62, and 15.47 respectively. These values indicate further reduction in errors compared to methods reported in other literature. Finally, the SAM-TCN prediction model is optimized using the Adam optimizer to improve the training effectiveness and convergence speed of the model. Experimental results demonstrate that the proposed method can effectively learn feature data, with prediction accuracy superior to other models.
- 2 September 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (17) , 7723
Open AccessMethod for Remaining Useful Life Prediction of Turbofan Engines Combining Adam Optimization-Based Self-Attention Mechanism with Temporal Convolutional Networks
Conducting the remaining useful life (RUL) prediction for an aircraft engines is of significant importance in enhancing aircraft operation safety and formulating reasonable maintenance plans. Addressing the issue of low prediction model accuracy due to traditional neural networks’ inability to fully extract key features, this paper proposes an engine RUL prediction model based on the adaptive moment estimation (Adam) optimized self-attention mechanism–temporal convolutional network (SAM-TCN) neural network. Firstly, the raw data monitored by sensors are normalized, and RUL labels are set. A sliding window is utilized for overlapping sampling of the data, capturing more temporal features while eliminating data dimensionality. Secondly, the SAM-TCN neural network prediction model is constructed. The temporal convolutional network (TCN) neural network is used to capture the temporal dependency between data, solving the mapping relationship of engine degradation characteristics. A self-attention mechanism (SAM) is employed to adaptively assign different weight contributions to different input features. In the experiments, the root mean square error (RMSE) values on four datasets are 11.50, 16.45, 11.62, and 15.47 respectively. These values indicate further reduction in errors compared to methods reported in other literature. Finally, the SAM-TCN prediction model is optimized using the Adam optimizer to improve the training effectiveness and convergence speed of the model. Experimental results demonstrate that the proposed method can effectively learn feature data, with prediction accuracy superior to other models.
- Open Access
A Study of Classroom Behavior Recognition Incorporating Super-Resolution and Target Detection
With the development of educational technology, machine learning and deep learning provide technical support for traditional classroom observation assessment. However, in real classroom scenarios, the technique faces challenges such as lack of clarity of raw images, complexity of datasets, multi-target detection errors, and complexity of character interactions. Based on the above problems, a student classroom behavior recognition network incorporating super-resolution and target detection is proposed. To cope with the problem of unclear original images in the classroom scenario, SRGAN (Super Resolution Generative Adversarial Network for Images) is used to improve the image resolution and thus the recognition accuracy. To address the dataset complexity and multi-targeting problems, feature extraction is optimized, and multi-scale feature recognition is enhanced by introducing AKConv and LASK attention mechanisms into the Backbone module of the YOLOv8s algorithm. To improve the character interaction complexity problem, the CBAM attention mechanism is integrated to enhance the recognition of important feature channels and spatial regions. Experiments show that it can detect six behaviors of students—raising their hands, reading, writing, playing on their cell phones, looking down, and leaning on the table—in high-definition images. And the accuracy and robustness of this network is verified. Compared with small-object detection algorithms such as Faster R-CNN, YOLOv5, and YOLOv8s, this network demonstrates good detection performance on low-resolution small objects, complex datasets with numerous targets, occlusion, and overlapping students.
Open AccessA Study of Classroom Behavior Recognition Incorporating Super-Resolution and Target Detection
With the development of educational technology, machine learning and deep learning provide technical support for traditional classroom observation assessment. However, in real classroom scenarios, the technique faces challenges such as lack of clarity of raw images, complexity of datasets, multi-target detection errors, and complexity of character interactions. Based on the above problems, a student classroom behavior recognition network incorporating super-resolution and target detection is proposed. To cope with the problem of unclear original images in the classroom scenario, SRGAN (Super Resolution Generative Adversarial Network for Images) is used to improve the image resolution and thus the recognition accuracy. To address the dataset complexity and multi-targeting problems, feature extraction is optimized, and multi-scale feature recognition is enhanced by introducing AKConv and LASK attention mechanisms into the Backbone module of the YOLOv8s algorithm. To improve the character interaction complexity problem, the CBAM attention mechanism is integrated to enhance the recognition of important feature channels and spatial regions. Experiments show that it can detect six behaviors of students—raising their hands, reading, writing, playing on their cell phones, looking down, and leaning on the table—in high-definition images. And the accuracy and robustness of this network is verified. Compared with small-object detection algorithms such as Faster R-CNN, YOLOv5, and YOLOv8s, this network demonstrates good detection performance on low-resolution small objects, complex datasets with numerous targets, occlusion, and overlapping students.
- Open Access
- 26 July 2024
- Published by MDPI AG in Preprints.org
A Study of Classroom Behavior Recognition Incorporating Super Resolution and Target Detection
With the development of educational technology, machine learning and deep learning provide technical support for traditional classroom observation assessment. However, in real classroom scenarios, the technique faces challenges such as lack of clarity of raw images, complexity of datasets, multi-target detection errors and complexity of character interactions.Based on the above problems, a student classroom behavior recognition network incorporating super-resolution and target detection is proposed. To cope with the problem of unclear original images in the classroom scenario, SRGAN (Super Resolution Generative Adversarial Network for Images) is used to improve the image resolution and thus the recognition accuracy.To address the dataset complexity and multi-targeting problems, feature extraction is optimized and multi-scale feature recognition is enhanced by introducing AKConv and LASK attention mechanisms into the Backbone module of YOLOv8s algorithm. To improve the character interaction complexity problem, the CBAM attention mechanism is integrated to enhance the recognition of important feature channels and spatial regions. Experiments show that it can detect six behaviors of students raising their hands, reading, writing, playing cell phones, looking down, and lying on the table in high-definition images. And the accuracy and robustness of this network are verified.Compared with small target detection algorithms such as Faster R-CNN, YOLOv5, and YOLOv8s, this network has better detection performance and can efficiently deal with the behavior recognition of multiple students.
- 26 July 2024
- Published by MDPI AG in Preprints.org
Open AccessA Study of Classroom Behavior Recognition Incorporating Super Resolution and Target Detection
With the development of educational technology, machine learning and deep learning provide technical support for traditional classroom observation assessment. However, in real classroom scenarios, the technique faces challenges such as lack of clarity of raw images, complexity of datasets, multi-target detection errors and complexity of character interactions.Based on the above problems, a student classroom behavior recognition network incorporating super-resolution and target detection is proposed. To cope with the problem of unclear original images in the classroom scenario, SRGAN (Super Resolution Generative Adversarial Network for Images) is used to improve the image resolution and thus the recognition accuracy.To address the dataset complexity and multi-targeting problems, feature extraction is optimized and multi-scale feature recognition is enhanced by introducing AKConv and LASK attention mechanisms into the Backbone module of YOLOv8s algorithm. To improve the character interaction complexity problem, the CBAM attention mechanism is integrated to enhance the recognition of important feature channels and spatial regions. Experiments show that it can detect six behaviors of students raising their hands, reading, writing, playing cell phones, looking down, and lying on the table in high-definition images. And the accuracy and robustness of this network are verified.Compared with small target detection algorithms such as Faster R-CNN, YOLOv5, and YOLOv8s, this network has better detection performance and can efficiently deal with the behavior recognition of multiple students.
- Open Access
- 25 March 2024
- Published by Wiley in International Journal of Aerospace Engineering
- Vol. 2024, 1-14
Aeroengine Remaining Life Prediction Using Feature Selection and Improved SE Blocks
Aeroengines use numerous sensors to detect equipment health and ensure proper operation. Currently, filtering useful sensor data and removing useless data is challenging in predicting the remaining useful life (RUL) of an aeroengine using deep learning. To reduce computational costs and improve prediction performance, we use random forest to evaluate the feature importance of sensor data. Based on the size of the feature corresponding to the Gini index, we select the appropriate sensor. This helps us to determine which sensor to use and ensures that the computational resources are not wasted on unnecessary sensors. Considering that the RUL of equipment changes in a progressively more complex manner as the equipment is used over time, we propose an improved squeeze and excitation block (SSE) and combine it with a convolutional neural network (CNN). By enhancing the feature selection ability of CNN through segmented squeeze and excitation block, the model can focus on important information within features to effectively improve prediction performance. We compared our experiments with other RUL experiments on the CMAPSS aeroengine dataset and then conducted ablation experiments to verify the critical role of the methods we used.
- 25 March 2024
- Published by Wiley in International Journal of Aerospace Engineering
- Vol. 2024, 1-14
Open AccessAeroengine Remaining Life Prediction Using Feature Selection and Improved SE Blocks
Aeroengines use numerous sensors to detect equipment health and ensure proper operation. Currently, filtering useful sensor data and removing useless data is challenging in predicting the remaining useful life (RUL) of an aeroengine using deep learning. To reduce computational costs and improve prediction performance, we use random forest to evaluate the feature importance of sensor data. Based on the size of the feature corresponding to the Gini index, we select the appropriate sensor. This helps us to determine which sensor to use and ensures that the computational resources are not wasted on unnecessary sensors. Considering that the RUL of equipment changes in a progressively more complex manner as the equipment is used over time, we propose an improved squeeze and excitation block (SSE) and combine it with a convolutional neural network (CNN). By enhancing the feature selection ability of CNN through segmented squeeze and excitation block, the model can focus on important information within features to effectively improve prediction performance. We compared our experiments with other RUL experiments on the CMAPSS aeroengine dataset and then conducted ablation experiments to verify the critical role of the methods we used.
- Open Access
- 28 February 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (5) , 1996
Turbofan Engine Health Prediction Model Based on ESO-BP Neural Network
Predicting the health index of turbofan engines is critical in reducing downtime and ensuring aircraft safety. This study introduces the elite snake optimizer-back propagation (ESO-BP) model to address the challenges of low accuracy and poor stability in predicting the health index of turbofan engines through neural networks. Firstly, the snake optimizer (SO) was improved into the elite snake optimizer (ESO) through an elite-guided strategy and a reverse learning mechanism. The performance improvement was validated using benchmark functions. Additionally, feature importance was introduced as a feature selection method. Finally, the optimization results of the ESO were employed to set the initial weights and biases of the BP neural network, preventing convergence to local optima. The prediction performance of the ESO-BP model was validated using the C-MAPSS datasets. The ESO-BP model was compared with the CNN, RNN, LSTM, baseline BP, and unimproved SO-BP models. The results demonstrated that the ESO-BP model has a superior accuracy with an impressive R-squared () value of 0.931 and a root mean square error () of 0.060 on the FD001 sub-dataset. Furthermore, the ESO-BP model exhibited lower standard deviations of evaluation metrics on 100 trials. According to the study, ESO-BP demonstrated a greater prediction accuracy and stability when compared to commonly used models such as CNN, RNN, LSTM, and BP.- 28 February 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (5) , 1996
Open AccessTurbofan Engine Health Prediction Model Based on ESO-BP Neural Network
Predicting the health index of turbofan engines is critical in reducing downtime and ensuring aircraft safety. This study introduces the elite snake optimizer-back propagation (ESO-BP) model to address the challenges of low accuracy and poor stability in predicting the health index of turbofan engines through neural networks. Firstly, the snake optimizer (SO) was improved into the elite snake optimizer (ESO) through an elite-guided strategy and a reverse learning mechanism. The performance improvement was validated using benchmark functions. Additionally, feature importance was introduced as a feature selection method. Finally, the optimization results of the ESO were employed to set the initial weights and biases of the BP neural network, preventing convergence to local optima. The prediction performance of the ESO-BP model was validated using the C-MAPSS datasets. The ESO-BP model was compared with the CNN, RNN, LSTM, baseline BP, and unimproved SO-BP models. The results demonstrated that the ESO-BP model has a superior accuracy with an impressive R-squared () value of 0.931 and a root mean square error () of 0.060 on the FD001 sub-dataset. Furthermore, the ESO-BP model exhibited lower standard deviations of evaluation metrics on 100 trials. According to the study, ESO-BP demonstrated a greater prediction accuracy and stability when compared to commonly used models such as CNN, RNN, LSTM, and BP.