MobileNet-CA-YOLO: An Improved YOLOv7 Based on the MobileNetV3 and Attention Mechanism for Rice Pests and Diseases Detection
Liangquan Jia, Tao Wang, Yi Chen, Ying Zang, Xiangge Li, Haojie Shi, Lu Gao
Abstract
The efficient identification of rice pests and diseases is crucial for preventing crop damage. To address the limitations of traditional manual detection methods and machine learning-based approaches, a new rice pest and disease recognition model based on an improved YOLOv7 algorithm has been developed. The model utilizes the lightweight network MobileNetV3 for feature extraction, reducing parameterization, and incorporates the coordinate attention mechanism (CA) and the SIoU loss function for enhanced accuracy. The model has been tested on a dataset of 3773 rice pest and disease images, achieving an accuracy of 92.3% and an [email protected] of 93.7%. The proposed MobileNet-CA-YOLO model is a high-performance and lightweight solution for rice pest and disease detection, providing accurate and timely results for farmers and researchers.