The Rice is a major food in India, playing an important role in the agricultural sector. However, various leaf diseases adversely affect rice production by reducing both quality and yield, leading to financial losses for farmers. Detecting these diseases at an early stage through automated methods can facilitate timely intervention and minimize crop damage. To classify diseases of rice, a novel method called ERROA-AlexNet is introduced. This model is designed to identify four categories of diseases such as bacterial leaf blight, rice leaf blast, brown spot, and tungro. The classification process utilizes AlexNet, with its weights optimized using the Enhanced Remora Rider Optimization Algorithm (ERROA), a hybrid approach that integrates the Enhanced Remora Optimization Algorithm (EROA) and Rider Optimization Algorithm (ROA). Experimental results, assessed utilizing a k-fold cross-validation technique, demonstrate that the proposed technique achieved an accuracy of 95.4%, a sensitivity of 94.3%, and a specificity of 98.1%. These results indicate that the ERROA-AlexNet approach outperforms conventional deep learning models such as RSW-Deep RNN, hybrid CNN-SVM and Deep CNN, as cited in the literature. This study focuses on the promising features of DL in precision agriculture, providing an efficient and reliable solution for automatic detection of rice leaf diseases.