A food source for billions of people worldwide, particularly in East and Southeast Asia, is rice. Unfortunately, several diseases, including bacterial blight, blast, brown spot, and tungro, can seriously reduce yields and often compromise the crop's ability to produce it. Using a dataset of 5,932 in-field images, this study explores the application of transfer learning to precisely identify and categorize these common rice leaf diseases. While pesticides may increase crop yields, they often degrade rice quality, which emphasizes the importance of early and accurate illness detection identification in order to assist farmers in minimizing losses. Using the most recent innovations in deep convolutional neural networks (DCNN), this study assesses the classification performance of five cutting-edge models: Xception, DenseNet121, InceptionResNetV2, ResNet50V2, and EfficientNetB0. The images underwent extensive pre-processing and data augmentation to enhance the quality and Variation of the training set, thereby improving model efficiency and generalization. With a 99.99% accuracy rate on the Mendeley dataset, the Xception model outperformed the other tested models. The results underscore the potential of transfer learning in agricultural disease diagnosis, offering a powerful tool for farmers to implement timely interventions and protect their crops from devastating losses.