A Modified Cross-Attention Deep Convolutional Neural Network for Early Detection of Rice Plant Diseases

Rice is the staple food crop grown in many countries around the world and demand for rice is highly increasing due to world population. Ensuring food security and environmental sustainability is very important since the unchecked plant might significantly affect crop yields. Several methods have been developed to detect rice plant disease but these methods are time-consuming, labor-intensive, and tend to human error. So, this research proposes a novel Modified Cross-Attention Deep Convolutional Neural Network (MCDCNN) model to enhance the accuracy and efficiency of rice plant disease detection. The Rice Leaf Disease Dataset and plant disease dataset are contributed to data collection in this work. The pre-processing methods significantly improve the quality of the collected rice plant images by enhancing contrast and eliminating noise. The proposed MCDCNN approach effectively captures the subtle visual cues indicative of diseases by using Deep Learning (DL) with the integration of the cross-attention mechanism. It improves the network’s focus on the most relevant features, improving detection accuracy. Experimental results are conducted to demonstrate the detection performance of the proposed approach. The experimental outcomes ascertain that the developed MCDCNN model attains a greater accuracy rate of 99.0% in both datasets, supporting improved early detection abilities that enable timely detection and crop losses. An ablation study is conducted to exhibit the contribution of each component within a proposed system. The performance of the developed MCDCNN demonstrates that it helps optimize resource management, support informed decision-making, and promote sustainable farming practices across diverse agricultural landscapes.