This research aims to develop a robust model for plant disease prediction in rice leaf images using the Regions with Convolutional Neural Networks (RCNN) approach. A comprehensive dataset comprising 17,000 images, including 400 images for each disease category (blast, brown spot, Hispa) and 100 images of healthy leaves, is collected and split into training and testing sets. Data augmentation techniques are employed to expand the dataset and improve model performance. Image annotation is performed to accurately label disease regions, enabling effective training and testing. The proposed RCNN architecture extracts features from regions of interest, followed by ROI pooling, classification, and regression processes. Activation and visualization techniques aid in understanding the model's decision-making process and feature extraction. Real-field images are used to test the model, and results demonstrate its accuracy in disease prediction and differentiation between healthy and diseased leaves. A confusion matrix analysis reveals high accuracies for brown spot (97.34%), Hispa (98.99%), healthy leaves (99%), and rice blast (97%). However, some misclassifications and challenges in distinguishing visually similar diseases are observed. These findings contribute to the advancement of automated plant disease prediction and facilitate early detection and intervention for effective disease management in rice crops. Future research can focus on refining the model to address misclassifications, exploring additional features, and incorporating transfer learning techniques to enhance accuracy and generalizability. The proposed approach holds significant potential in supporting farmers and agronomists in improving crop productivity, reducing yield losses, and promoting sustainable farming practices. Overall, this research provides valuable insights and a reliable framework for plant disease prediction in rice leaf images, contributing to the field of agricultural technology.