Lei Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Lei Zhu has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Deming Ma
College of Metrology Measurement and Instrument, China Jiliang University, Hangzhou 310018, China
1 shared publication - Jiahao Duan
College of Metrology Measurement and Instrument, China Jiliang University, Hangzhou 310018, China
1 shared publication - Lu Liu
China Jiliang University
1 shared publication - Ming Kong
College of Metrology Measurement and Instrument, China Jiliang University, Hangzhou, China
1 shared publication - Shiling Wang
College of Metrology Measurement and Instrument, China Jiliang University, Hangzhou 310018, China
1 shared publication
Latest Publications
- Open Access
- 1 April 2026
- Published by MDPI AG in Algorithms
- Vol. 19 (4) , 267
LW-YOLO: A Lightweight Enhanced YOLOv8-Based Model for Mosquito Detection
Mosquitoes of the genera Aedes and Culex are major vectors of mosquito-borne diseases, posing serious threats to public health. Accurate detection of these species is therefore crucial for disease prevention and vector control. Traditional identification methods are time-consuming, labor-intensive, and prone to human error. With the rapid development of deep learning, automated mosquito detection has become feasible; however, existing object detection models still struggle with small-object recognition and high computational complexity. To address these limitations, this study constructs a self-developed dataset and proposes a lightweight mosquito detection model based on YOLOv8, termed LW-YOLO. The model integrates HGNetv2, Rep-Ghost, and SCDH modules into the backbone, neck, and head, respectively, enhancing both detection accuracy and computational efficiency. Experimental results show that LW-YOLO achieves a precision of 0.978, recall of 0.972, and mAP50 of 0.987, improving by 1.6%, 1.25%, and 0.7% over the baseline YOLOv8. Meanwhile, its parameter count and computational cost are reduced from 3.0 M and 8.1 GFLOPs to 1.2 M and 4.4 GFLOPs, corresponding to decreases of 60% and 45.7%, respectively. The proposed LW-YOLO model not only achieves accurate detection of Aedes and Culex mosquitoes, providing technical support for mosquito-borne disease prevention, but also offers a promising lightweight solution for deployment on resource-constrained embedded or edge devices.
- 1 April 2026
- Published by MDPI AG in Algorithms
- Vol. 19 (4) , 267
Open AccessLW-YOLO: A Lightweight Enhanced YOLOv8-Based Model for Mosquito Detection
Mosquitoes of the genera Aedes and Culex are major vectors of mosquito-borne diseases, posing serious threats to public health. Accurate detection of these species is therefore crucial for disease prevention and vector control. Traditional identification methods are time-consuming, labor-intensive, and prone to human error. With the rapid development of deep learning, automated mosquito detection has become feasible; however, existing object detection models still struggle with small-object recognition and high computational complexity. To address these limitations, this study constructs a self-developed dataset and proposes a lightweight mosquito detection model based on YOLOv8, termed LW-YOLO. The model integrates HGNetv2, Rep-Ghost, and SCDH modules into the backbone, neck, and head, respectively, enhancing both detection accuracy and computational efficiency. Experimental results show that LW-YOLO achieves a precision of 0.978, recall of 0.972, and mAP50 of 0.987, improving by 1.6%, 1.25%, and 0.7% over the baseline YOLOv8. Meanwhile, its parameter count and computational cost are reduced from 3.0 M and 8.1 GFLOPs to 1.2 M and 4.4 GFLOPs, corresponding to decreases of 60% and 45.7%, respectively. The proposed LW-YOLO model not only achieves accurate detection of Aedes and Culex mosquitoes, providing technical support for mosquito-borne disease prevention, but also offers a promising lightweight solution for deployment on resource-constrained embedded or edge devices.