Abstract

The objective of this study was to develop a real-time application for paddy disease classification using 2-Dimensional Convolutional Neural Network (CNN) Faster RCNN and ResNet50. It presents a system developed to classify three prevalent paddy diseases namely bacterial blight, tungro, and brown spot. The system involves three primary steps: dataset collection, training, and testing. Real-time images of paddy plants were utilized. The dataset for paddy disease classification comprised of 2,296 images for training and 575 images for testing, with each image labeled according to the particular disease it depicts. The Convolutional Neural Network (CNN), Faster RCNN and ResNet50 models were employed to classify paddy diseases based on the collected images. Experimental results indicated that ResNet50 demonstrates superior performance, achieving an accuracy of 99.43% in disease classification compared to Faster RCNN and the basic 2D-CNN models. This demonstrated ResNet50's effectiveness in distinguishing between tungro, bacterial blight, and brown spot in paddy plants. The application utilizes a specifically curated paddy disease dataset and implements advanced deep learning techniques for real-time classification. The system's high accuracy and efficiency in detecting paddy diseases in various field conditions represent a significant advancement in agricultural technology.