ENHANCING PRECISION AGRICULTURE THROUGH ATTENTION-BASED DEEP LEARNING FOR PADDY FIELD SEGMENTATION USING ATTNETS

Precision agriculture is revolutionizing farming by leveraging technology for efficient resource management and higher crop yields. A significant challenge in this domain is accurate segmentation of paddy fields from aerial and satellite imagery to monitor crop health and optimize farming practices. This study introduces an Attention Network (AttNet) framework designed specifically for paddy field segmentation. AttNet incorporates attention mechanisms to enhance spatial and contextual feature extraction, improving segmentation accuracy over conventional models. The proposed method was trained and tested on a dataset of satellite images, achieving a mean Intersection over Union (mIoU) of 91.5% and pixel accuracy of 94.2%, surpassing state-of-the-art methods such as U-Net (mIoU: 86.3%) and DeepLabv3 (mIoU: 88.7%). This demonstrates its effectiveness in handling complex paddy field patterns. The results validate AttNet as a powerful tool for agricultural applications, promising to aid farmers and policymakers in making informed decisions. Future research will explore extending this framework to other crop types and integrating it with real-time UAV systems.