Fufang Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Fufang Li has a total of 58 co-authors. Here are the top co-authors by number of shared publications:
Yuanyong Feng
School of Computer Science and Cyber Engineering, Guangzhou University, Guangzhou 510006, China
4 shared publicationsShaobo Zhang
School of Computer Science and Technology, Hunan University of Science and Technology, Xiangtan, China
3 shared publications- Guosheng Huang
School of Information Science and Engineering, Hunan First Normal University, Changsha, China
2 shared publications Jinrong Zhang
Guangzhou University
2 shared publicationsTian Wang
College of Computer Science and Technology, Huaqiao University, Xiamen, China
2 shared publicationsWeihao Lu
Guangzhou University
2 shared publications
Latest Publications
- Open Access
- 16 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16  (1)
A novel hybrid model of simplified and external attention coupled with enhanced CNN for medical image segmentation
Although UNet has proven its success in various tasks involving medical image segmentation, its capacity to capture global context is restricted by the finite receptive field inherent to convolutional operations. Transformer is capable of capturing long-range dependencies. Consequently, integrating transformer into UNet can alleviate the issue of its limited receptive field. However, transformer usually relies heavily on large-scale pre-training and struggles to capture local features. To address these challenges, we propose SimEANet, a network that employs an encoder-decoder structure with a hybrid CNN-Transformer architecture. We design an enhanced ResNet as a shallow feature extractor for the encoder. Furthermore, we introduce SimEA transformer as the backbone of the encoder. Finally, we use improved cascaded upsampling processors to obtain the segmentation result. The performance of SimEANet is substantiated through rigorous testing on two public accessible datasets. Extensive experiments demonstrate the high competitiveness of our approach, achieving average Dice Similarity Coefficients (DSC) of 82.35% and 91.85% on two datasets. SimEANet notably enhances performance in multi-organ segmentation tasks, achieving an advanced level of segmentation accuracy.
- 16 March 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16  (1)
Open AccessA novel hybrid model of simplified and external attention coupled with enhanced CNN for medical image segmentation
Although UNet has proven its success in various tasks involving medical image segmentation, its capacity to capture global context is restricted by the finite receptive field inherent to convolutional operations. Transformer is capable of capturing long-range dependencies. Consequently, integrating transformer into UNet can alleviate the issue of its limited receptive field. However, transformer usually relies heavily on large-scale pre-training and struggles to capture local features. To address these challenges, we propose SimEANet, a network that employs an encoder-decoder structure with a hybrid CNN-Transformer architecture. We design an enhanced ResNet as a shallow feature extractor for the encoder. Furthermore, we introduce SimEA transformer as the backbone of the encoder. Finally, we use improved cascaded upsampling processors to obtain the segmentation result. The performance of SimEANet is substantiated through rigorous testing on two public accessible datasets. Extensive experiments demonstrate the high competitiveness of our approach, achieving average Dice Similarity Coefficients (DSC) of 82.35% and 91.85% on two datasets. SimEANet notably enhances performance in multi-organ segmentation tasks, achieving an advanced level of segmentation accuracy.
- 15 January 2026
- Published by Springer Nature
- p. 335-350
Bridging Domain Shifts with 1–5 Shots: Unified Elastic Prototype-Contrastive Learning for Source-Free Hashing Adaptation
- 15 January 2026
- Published by Springer Nature
- p. 335-350
Bridging Domain Shifts with 1–5 Shots: Unified Elastic Prototype-Contrastive Learning for Source-Free Hashing Adaptation
- 14 November 2025
- Published by Springer Nature
- p. 561-575
PCSR: A Two-Stage Prompt-Guided and Semantic Refinement Framework for CLIP-Based Few-Shot Learning
- 14 November 2025
- Published by Springer Nature
- p. 561-575
PCSR: A Two-Stage Prompt-Guided and Semantic Refinement Framework for CLIP-Based Few-Shot Learning
- Open Access
- 13 November 2025
- Published by MDPI AG in Biomimetics
- Vol. 10  (11) , 770
EABI-DETR: An Efficient Aerial Small Object Detection Network
Small object detection, as an important research topic in computer vision, has been widely applied in aerial visual tasks such as remote sensing and UAV imagery. However, due to challenges such as small object size, large-scale variations, and complex backgrounds, existing detection models often struggle to capture fine-grained semantics and high-resolution texture information in aerial scenes, leading to limited performance. To address these issues, this paper proposes an efficient aerial small object detection model, EABI-DETR (Efficient Attention and Bi-level Integration DETR), based on the RT-DETR framework. The proposed model introduces systematic enhancements from three aspects: (1) A lightweight backbone network, C2f-EMA, is developed by integrating the C2f structure with an efficient multi-scale attention (EMA) mechanism. This design jointly models channel semantics and spatial details with minimal computational overhead, thereby strengthening the perception of small objects. (2) A P2-BiFPN bi-directional multi-scale fusion module is further designed to incorporate shallow high-resolution features. Through top-down and bottom-up feature interactions, this module enhances cross-scale information flow and effectively preserves the fine details and textures of small objects. (3) To improve localization robustness, a Focaler-MPDIoU loss function is introduced to better handle hard samples during regression optimization. Experiments conducted on the VisDrone2019 dataset demonstrate that EABI-DETR achieves 53.4% mAP@0.5 and 34.1% mAP@0.5:0.95, outperforming RT-DETR by 6.2% and 5.1%, respectively, while maintaining high inference efficiency. These results confirm the effectiveness of integrating lightweight attention mechanisms and shallow feature fusion for aerial small object detection, offering a new paradigm for efficient UAV-based visual perception.
- 13 November 2025
- Published by MDPI AG in Biomimetics
- Vol. 10  (11) , 770
Open AccessEABI-DETR: An Efficient Aerial Small Object Detection Network
Small object detection, as an important research topic in computer vision, has been widely applied in aerial visual tasks such as remote sensing and UAV imagery. However, due to challenges such as small object size, large-scale variations, and complex backgrounds, existing detection models often struggle to capture fine-grained semantics and high-resolution texture information in aerial scenes, leading to limited performance. To address these issues, this paper proposes an efficient aerial small object detection model, EABI-DETR (Efficient Attention and Bi-level Integration DETR), based on the RT-DETR framework. The proposed model introduces systematic enhancements from three aspects: (1) A lightweight backbone network, C2f-EMA, is developed by integrating the C2f structure with an efficient multi-scale attention (EMA) mechanism. This design jointly models channel semantics and spatial details with minimal computational overhead, thereby strengthening the perception of small objects. (2) A P2-BiFPN bi-directional multi-scale fusion module is further designed to incorporate shallow high-resolution features. Through top-down and bottom-up feature interactions, this module enhances cross-scale information flow and effectively preserves the fine details and textures of small objects. (3) To improve localization robustness, a Focaler-MPDIoU loss function is introduced to better handle hard samples during regression optimization. Experiments conducted on the VisDrone2019 dataset demonstrate that EABI-DETR achieves 53.4% mAP@0.5 and 34.1% mAP@0.5:0.95, outperforming RT-DETR by 6.2% and 5.1%, respectively, while maintaining high inference efficiency. These results confirm the effectiveness of integrating lightweight attention mechanisms and shallow feature fusion for aerial small object detection, offering a new paradigm for efficient UAV-based visual perception.
- Open Access
- 17 October 2025
- Published by Angus Academic Press in Advances in Computer and Materials Scienc Research
- Vol. 4  (1) , 61-68
CSSWin-UNet: A Transformer UNet Combined with Patch-Based and Cross-Shaped Windows for Medical Image Segmentation
In medical image segmentation, both Convolutional Neural Networks (CNNs) and self-attention mechanisms have shown success but also have limitations. CNNs excel at capturing local features but struggle with long-range dependencies, while self-attention effectively models global context but is computationally intensive and may lose fine local details. To address these challenges, we propose CSSWin-UNet, a novel U-shaped architecture that integrates two complementary self-attention mechanisms(Swin and CSWin Transformer) to achieve a balanced representation of both local features and global dependencies, while substantially reducing computational overhead. Experimental results on multiple public benchmark datasets demonstrate that CSSWin-UNet delivers superior segmentation performance with significantly lower model complexity, highlighting its potential for practical deployment in medical imaging tasks.
- 17 October 2025
- Published by Angus Academic Press in Advances in Computer and Materials Scienc Research
- Vol. 4  (1) , 61-68
Open AccessCSSWin-UNet: A Transformer UNet Combined with Patch-Based and Cross-Shaped Windows for Medical Image Segmentation
In medical image segmentation, both Convolutional Neural Networks (CNNs) and self-attention mechanisms have shown success but also have limitations. CNNs excel at capturing local features but struggle with long-range dependencies, while self-attention effectively models global context but is computationally intensive and may lose fine local details. To address these challenges, we propose CSSWin-UNet, a novel U-shaped architecture that integrates two complementary self-attention mechanisms(Swin and CSWin Transformer) to achieve a balanced representation of both local features and global dependencies, while substantially reducing computational overhead. Experimental results on multiple public benchmark datasets demonstrate that CSSWin-UNet delivers superior segmentation performance with significantly lower model complexity, highlighting its potential for practical deployment in medical imaging tasks.
- Open Access
- 1 January 2025
- Published by MDPI AG in Biomimetics
- Vol. 10  (1) , 16
EMNet: A Novel Few-Shot Image Classification Model with Enhanced Self-Correlation Attention and Multi-Branch Joint Module
In this research, inspired by the principles of biological visual attention mechanisms and swarm intelligence found in nature, we present an Enhanced Self-Correlation Attention and Multi-Branch Joint Module Network (EMNet), a novel model for few-shot image classification. Few-shot image classification aims to address the problem of image classification when data are limited. Traditional models require a large amount of labeled data for training, while few-shot learning trains models using only a small number of samples (just a few samples per class) to recognize new categories. EMNet shows its potential for bio-inspired algorithms in optimizing feature extraction and enhancing generalization capabilities. It features two key modules: Enhanced Self-Correlated Attention (ESCA) and Multi-Branch Joint Module (MBJ Module). EMNet tackles two main challenges in few-shot learning: how to make an effective important feature extraction and enhancement in images, and improving generalization to new categories. The ESCA module boosts the precision in extracting crucial local features, enhancing classification accuracy. The MBJ module focuses on shared features across images, emphasizing similarities within classes and subtle differences between them. This enhances model adaptability and generalization to new categories. Experimental results show that our model performs better than existing models in one-shot and five-shot tasks on mini-ImageNet, CUB-200, and CIFAR-FS datasets, which proves the proposed model to be an efficient end-to-end solution for few-shot image classification. In the five-way one-shot and five-way five-shot experiments on the CUB-200-2011 dataset, EMNet achieved classification accuracies that were 1.27 and 0.54 percentage points higher than those of RENet, respectively. In the five-way one-shot and five-way five-shot experiments on the miniImageNet dataset, EMNet’s classification accuracies were 0.02 and 0.48 percentage points higher than those of RENet, respectively. In the five-way one-shot and five-way five-shot experiments on the CIFAR-FS dataset, EMNet’s classification accuracies were 0.19 and 0.18 percentage points higher than those of RENet.
- 1 January 2025
- Published by MDPI AG in Biomimetics
- Vol. 10  (1) , 16
Open AccessEMNet: A Novel Few-Shot Image Classification Model with Enhanced Self-Correlation Attention and Multi-Branch Joint Module
In this research, inspired by the principles of biological visual attention mechanisms and swarm intelligence found in nature, we present an Enhanced Self-Correlation Attention and Multi-Branch Joint Module Network (EMNet), a novel model for few-shot image classification. Few-shot image classification aims to address the problem of image classification when data are limited. Traditional models require a large amount of labeled data for training, while few-shot learning trains models using only a small number of samples (just a few samples per class) to recognize new categories. EMNet shows its potential for bio-inspired algorithms in optimizing feature extraction and enhancing generalization capabilities. It features two key modules: Enhanced Self-Correlated Attention (ESCA) and Multi-Branch Joint Module (MBJ Module). EMNet tackles two main challenges in few-shot learning: how to make an effective important feature extraction and enhancement in images, and improving generalization to new categories. The ESCA module boosts the precision in extracting crucial local features, enhancing classification accuracy. The MBJ module focuses on shared features across images, emphasizing similarities within classes and subtle differences between them. This enhances model adaptability and generalization to new categories. Experimental results show that our model performs better than existing models in one-shot and five-shot tasks on mini-ImageNet, CUB-200, and CIFAR-FS datasets, which proves the proposed model to be an efficient end-to-end solution for few-shot image classification. In the five-way one-shot and five-way five-shot experiments on the CUB-200-2011 dataset, EMNet achieved classification accuracies that were 1.27 and 0.54 percentage points higher than those of RENet, respectively. In the five-way one-shot and five-way five-shot experiments on the miniImageNet dataset, EMNet’s classification accuracies were 0.02 and 0.48 percentage points higher than those of RENet, respectively. In the five-way one-shot and five-way five-shot experiments on the CIFAR-FS dataset, EMNet’s classification accuracies were 0.19 and 0.18 percentage points higher than those of RENet.
- Open Access
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 323-333
Domain Alignment with Large Vision-language Models for Cross-domain Remote Sensing Image Retrieval
Cross-domain remote sensing image retrieval has been a hotspot in the past few years. Most of the existing methods focus on combining semantic learning with domain adaptation on well-labeled source domain and unlabeled target domain. However, they face two serious challenges. (1) They cannot deal with practical scenarios where the source domain lacks sufficient label supervision. (2) They suffer from severe performance degradation when the data distribution between the source domain and target domain becomes highly inconsistent. To address these challenges, we propose D omain A lignment with L arge V ision-language models for cross-domain remote sensing image retrieval (termed as DALV). First, we design a dual-modality prototype guided pseudo-labeling mechanism, which leverages the pre-trained large vision-language model (i.e., CLIP) to assign pseudo-labels for all unlabeled source domain images and target domain images. Second, we compute the confidence scores for these pseudo-labels to distinguish their reliability. Next, we devise a loss reweighting strategy, which incorporates the confidence scores as weight values into the contrastive loss to mitigate the impact of noisy pseudo-labels. Finally, the low-rank adaptation fine-tuning means is adapted to update our model and achieve domain alignment to obtain class discriminative features. Extensive experiments on 12 cross-domain remote sensing image retrieval tasks show that our proposed DALV outperforms the state-of-the-art approaches. The source code is available at https://github.com/ptyy01/DALV.
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 323-333
Open AccessDomain Alignment with Large Vision-language Models for Cross-domain Remote Sensing Image Retrieval
Cross-domain remote sensing image retrieval has been a hotspot in the past few years. Most of the existing methods focus on combining semantic learning with domain adaptation on well-labeled source domain and unlabeled target domain. However, they face two serious challenges. (1) They cannot deal with practical scenarios where the source domain lacks sufficient label supervision. (2) They suffer from severe performance degradation when the data distribution between the source domain and target domain becomes highly inconsistent. To address these challenges, we propose D omain A lignment with L arge V ision-language models for cross-domain remote sensing image retrieval (termed as DALV). First, we design a dual-modality prototype guided pseudo-labeling mechanism, which leverages the pre-trained large vision-language model (i.e., CLIP) to assign pseudo-labels for all unlabeled source domain images and target domain images. Second, we compute the confidence scores for these pseudo-labels to distinguish their reliability. Next, we devise a loss reweighting strategy, which incorporates the confidence scores as weight values into the contrastive loss to mitigate the impact of noisy pseudo-labels. Finally, the low-rank adaptation fine-tuning means is adapted to update our model and achieve domain alignment to obtain class discriminative features. Extensive experiments on 12 cross-domain remote sensing image retrieval tasks show that our proposed DALV outperforms the state-of-the-art approaches. The source code is available at https://github.com/ptyy01/DALV.
- Open Access
- 14 July 2024
- Published by MDPI AG in Mathematics
- Vol. 12  (14) , 2206
Algorithms for Densest Subgraphs of Vertex-Weighted Graphs
Finding the densest subgraph has tremendous potential in computer vision and social network research, among other domains. In computer vision, it can demonstrate essential structures, and in social network research, it aids in identifying closely associated communities. The densest subgraph problem is finding a subgraph with maximum mean density. However, most densest subgraph-finding algorithms are based on edge-weighted graphs, where edge weights can only represent a single value dimension, whereas practical applications involve multiple dimensions. To resolve the challenge, we propose two algorithms for resolving the densest subgraph problem in a vertex-weighted graph. First, we present an exact algorithm that builds upon Goldberg’s original algorithm. Through theoretical exploration and analysis, we rigorously verify our proposed algorithm’s correctness and confirm that it can efficiently run in polynomial time O(n(n + m)log2n) is its temporal complexity. Our approach can be applied to identify closely related subgroups demonstrating the maximum average density in real-life situations. Additionally, we consistently offer an approximation algorithm that guarantees an accurate approximation ratio of 2. In conclusion, our contributions enrich theoretical foundations for addressing the densest subgraph problem.
- 14 July 2024
- Published by MDPI AG in Mathematics
- Vol. 12  (14) , 2206
Open AccessAlgorithms for Densest Subgraphs of Vertex-Weighted Graphs
Finding the densest subgraph has tremendous potential in computer vision and social network research, among other domains. In computer vision, it can demonstrate essential structures, and in social network research, it aids in identifying closely associated communities. The densest subgraph problem is finding a subgraph with maximum mean density. However, most densest subgraph-finding algorithms are based on edge-weighted graphs, where edge weights can only represent a single value dimension, whereas practical applications involve multiple dimensions. To resolve the challenge, we propose two algorithms for resolving the densest subgraph problem in a vertex-weighted graph. First, we present an exact algorithm that builds upon Goldberg’s original algorithm. Through theoretical exploration and analysis, we rigorously verify our proposed algorithm’s correctness and confirm that it can efficiently run in polynomial time O(n(n + m)log2n) is its temporal complexity. Our approach can be applied to identify closely related subgroups demonstrating the maximum average density in real-life situations. Additionally, we consistently offer an approximation algorithm that guarantees an accurate approximation ratio of 2. In conclusion, our contributions enrich theoretical foundations for addressing the densest subgraph problem.
- Open Access
Helmet-Wearing Tracking Detection Based on StrongSORT
Object detection based on deep learning is one of the most important and fundamental tasks of computer vision. High-performance detection algorithms have been widely used in many practical fields. For the management of workers wearing helmets in construction scenarios, this paper proposes a framework model based on the YOLOv5 detection algorithm, combined with multi-object tracking algorithms, to monitor and track whether workers wear safety helmets in real-time video. The improved StrongSORT tracking algorithm of DeepSORT is selected to reduce the loss of the tracked object caused by the occlusion, trajectory blur, and motion scale of the object. The safety helmet dataset is trained with YOLOv5s, and the best result of training is used as the weight model in the StrongSORT tracking algorithm. The experimental results show that the mAP@0.5 of all classes in the YOLOv5s model can reach 95.1% in the validation dataset, mAP@0.5:0.95 is 62.1%, and the precision of wearing helmet is 95.7%. After the box regression loss function was changed from CIOU to Focal-EIOU, the mAP@0.5 increased to 95.4%, mAP@0.5:0.95 increased to 62.9%, and the precision of wearing helmet increased to 96.5%, which were increased by 0.3%, 0.8% and 0.8%, respectively. StrongSORT can update object trajectories in video frames at a speed of 0.05 s per frame. Based on the improved YOLOv5s combined with the StrongSORT tracking algorithm, the helmet-wearing tracking detection can achieve better performance.
Open AccessHelmet-Wearing Tracking Detection Based on StrongSORT
Object detection based on deep learning is one of the most important and fundamental tasks of computer vision. High-performance detection algorithms have been widely used in many practical fields. For the management of workers wearing helmets in construction scenarios, this paper proposes a framework model based on the YOLOv5 detection algorithm, combined with multi-object tracking algorithms, to monitor and track whether workers wear safety helmets in real-time video. The improved StrongSORT tracking algorithm of DeepSORT is selected to reduce the loss of the tracked object caused by the occlusion, trajectory blur, and motion scale of the object. The safety helmet dataset is trained with YOLOv5s, and the best result of training is used as the weight model in the StrongSORT tracking algorithm. The experimental results show that the mAP@0.5 of all classes in the YOLOv5s model can reach 95.1% in the validation dataset, mAP@0.5:0.95 is 62.1%, and the precision of wearing helmet is 95.7%. After the box regression loss function was changed from CIOU to Focal-EIOU, the mAP@0.5 increased to 95.4%, mAP@0.5:0.95 increased to 62.9%, and the precision of wearing helmet increased to 96.5%, which were increased by 0.3%, 0.8% and 0.8%, respectively. StrongSORT can update object trajectories in video frames at a speed of 0.05 s per frame. Based on the improved YOLOv5s combined with the StrongSORT tracking algorithm, the helmet-wearing tracking detection can achieve better performance.
- Open Access
- 1 April 2021
- Published by SAGE Publications in International Journal of Distributed Sensor Networks
- Vol. 17  (4)
Parameter analysis and optimization of polling-based medium access control protocol for multi-sensor communication
In this article, a comprehensive parameter analysis for the polling-based medium access control protocol is executed. The theoretical expressions of the relationship between network parameters and performance including delay and energy consumption are given for the first time. The specific conclusions in this article are as follows: (1) awake duration is the parameter that has the greatest impact on delay and energy consumption. Increasing the duty cycle (awake duration) will effectively reduce the delay, but will also increase the energy consumption within a certain range; (2) increasing polling duration can reduce the delay, but it will also increase the energy consumption; and (3) more forwarding nodes cause a smaller delay, and it can save the energy with modest increase of delay by reducing the polling duration. An adaptive parameter optimization polling-based medium access control protocol is proposed to optimize network performance. In this protocol, the residual energy gets fully used to increase awake duration and polling duration, which makes the delay smaller, and the network maintains a long lifetime meanwhile. Based on the results of the analysis, the adaptive parameter optimization polling-based medium access control protocol proposed in this article reduces the delay by 22.40% and increases the energy efficiency by 23.25%.
- 1 April 2021
- Published by SAGE Publications in International Journal of Distributed Sensor Networks
- Vol. 17  (4)
Open AccessParameter analysis and optimization of polling-based medium access control protocol for multi-sensor communication
In this article, a comprehensive parameter analysis for the polling-based medium access control protocol is executed. The theoretical expressions of the relationship between network parameters and performance including delay and energy consumption are given for the first time. The specific conclusions in this article are as follows: (1) awake duration is the parameter that has the greatest impact on delay and energy consumption. Increasing the duty cycle (awake duration) will effectively reduce the delay, but will also increase the energy consumption within a certain range; (2) increasing polling duration can reduce the delay, but it will also increase the energy consumption; and (3) more forwarding nodes cause a smaller delay, and it can save the energy with modest increase of delay by reducing the polling duration. An adaptive parameter optimization polling-based medium access control protocol is proposed to optimize network performance. In this protocol, the residual energy gets fully used to increase awake duration and polling duration, which makes the delay smaller, and the network maintains a long lifetime meanwhile. Based on the results of the analysis, the adaptive parameter optimization polling-based medium access control protocol proposed in this article reduces the delay by 22.40% and increases the energy efficiency by 23.25%.