Publication

Abstract : The agricultural sector serves as a cornerstone in the socioeconomic landscape of nations worldwide, with paddy rice standing as a vital staple crop in many regions. However, the proliferation of common paddy leaf diseases presents significant challenges to the global quality and quantity of rice crop yields. Early detection of these diseases is imperative to mitigate their impact on crop production. Leveraging future-tech UAV s (Unmanned Aerial Vehicles) network for remote sensing coupled with Deep Learning (DL) holds promise in addressing this issue. This paper introduces Tiny-YOLOv9, a novel lightweight architecture derived from YOLOv9, explicitly tailored for real- time leaf disease detection across various plant species. Tiny-YOLOv9 integrates cutting-edge components such as the 3D Fea-ture Adaptation Module (3D-FAM), Deep Wise Point Convolution (DWC), Coordinate Attention Module (CAM), and Convolutional Block Attention Modules (CBAM) to enhance feature extraction precision and attention. The proposed methodology exhibits superior performance and detection capabilities compared to the state-of-the-art (SOTA), as evidenced by metrics such as Average Precision (AP), Average recall (AR), Fl-Score, and mean Average Precision (mAP). The minimal resource utilization and enhanced detection accuracy make the proposed Tiny- YOLOv9 a better alternative for UAV (Unmanned Arial Vehicles) onboard intelligence for paddy agronomy.