Comparative Analysis of Deep Learning Architectures and Optimizers for Paddy Leaf Disease Classification

Rice, an essential food item on a global scale, faces obstacles to production that are caused by diseases, specifically leaf diseases, emphasising the need for effective detection techniques. The objective of this research is to present an automated system that can diagnose maladies affecting paddy leaves. The system is specifically developed to tackle the challenges associated with manual identification, which are particularly prevalent in remote regions where expert resources are limited. The methodology entails the application of AlexNet and SqueezeNet, which are two discrete deep learning approaches. AlexNet, an advanced classification methodology, is trained using a comprehensive dataset and transfer learning strategies are implemented to improve its ability to identify intricate patterns that are commonly associated with rice leaf diseases such as bacterial blight, brown spot, leaf blast, and leaf smut. In contrast, SqueezeNet, which is specifically designed for the purpose of efficiently classifying binary images, utilizes binary neural networks in order to reduce computational complexity and maintain competitive performance. The assessment of the system's performance indicates that both SqueezeNet and AlexNet are effective in precisely diagnosing and categorising diseases affecting maize leaves.