<p>Rice has an important place as food for more than half of the world’s population, but its yield and stability are severely affected by rice diseases. Current rice disease detection methods are inefficient and costly. To address this issue, this study collected 1505 data points and images from rice fields and the web and annotated them for training frameworks. After that, this study developed a new rice detection framework, MFAC-YOLOv8, based on the YOLOv8 network. The framework integrates the MobileNetv4 network and the Focal Modulation module and uses them as the backbone network of the improved YOLOv8 to improve the detection accuracy of the network. In addition, AKConv and the Context Guided block are introduced and used to further improve the neck network of YOLOv8, which simplifies the framework and further enhances the detection. The experimental results show that the MFAC-YOLOv8 framework exhibits excellent performance in all evaluation metrics, with 8.1<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11554_2025_1661_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">$\%$</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, 2.2<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11554_2025_1661_Article_IEq2.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">$\%$</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, and 3.4<InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11554_2025_1661_Article_IEq3.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">$\%$</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation> improvements in accuracy, recall, and mean average precision, respectively, compared with the baseline framework. In addition, the framework achieves a high frame rate of 166 frames per second (FPS) and a relatively compact parameter size of 18.4 million. These results suggest that the proposed method has great potential for effective rice disease detection.</p>

Identification of rice disease based on MFAC-YOLOv8

期刊论文
发表日期：

2025年3月10日

查看全文

Bingyang Wang,
Huibo Zhou,
Hui Xie,
Ruolan Chen

摘要

Rice has an important place as food for more than half of the world’s population, but its yield and stability are severely affected by rice diseases. Current rice disease detection methods are inefficient and costly. To address this issue, this study collected 1505 data points and images from rice fields and the web and annotated them for training frameworks. After that, this study developed a new rice detection framework, MFAC-YOLOv8, based on the YOLOv8 network. The framework integrates the MobileNetv4 network and the Focal Modulation module and uses them as the backbone network of the improved YOLOv8 to improve the detection accuracy of the network. In addition, AKConv and the Context Guided block are introduced and used to further improve the neck network of YOLOv8, which simplifies the framework and further enhances the detection. The experimental results show that the MFAC-YOLOv8 framework exhibits excellent performance in all evaluation metrics, with 8.1 $\%$ $%$ , 2.2 $\%$ $%$ , and 3.4 $\%$ $%$ improvements in accuracy, recall, and mean average precision, respectively, compared with the baseline framework. In addition, the framework achieves a high frame rate of 166 frames per second (FPS) and a relatively compact parameter size of 18.4 million. These results suggest that the proposed method has great potential for effective rice disease detection.