Rice is commonly found on every Indian plate, either as a dish or as an essential ingredient, but its agriculture faces significant challenges due to various diseases affecting rice leaves. There are around 19 rice leaf diseases, and this study focused on three commonly observed types: brown spot, leaf smut, and bacterial leaf blight. The objective was to build a deep learning model to identify rice leaf disease at its early stages and create an interface where users can input rice leaf images to identify any disease present, allowing action to be taken before it spreads and preventing yield loss. For this purpose, the “Rice Leaf Diseases Dataset” containing 120 images of disease-affected rice leaves, is taken for experimental work. Through augmentation, the dataset was expanded, and pictures were preprocessed. Three transfer learning architectures—ResNet 18, ResNet 50, and AlexNet—were tested, with ResNet18 performing best, achieving 95.47% accuracy and a validation loss of 0.1376. In the future, adding more real-world datasets could further enhance model performance, improving accuracy and reliability in detecting rice leaf diseases.