Comparative Analysis of Deep Learning Approaches for Enhanced Detection and Classification of Paddy Plant Diseases

Abstract

Paddy is among the most relevant staple crops in the world and numerous diseases that impact its production largely. Early and proper disease diagnosis of the paddy plants is necessary to avoid the loss of crops and enhance agricultural production. As artificial intelligence (AI) continues to develop, deep learning (DL) methods are proving to be advantageous in automated identification of plant diseases by analyzing images. In this work, a comparative study on the various approaches of DL is provided in order to better detect and classify paddy plant diseases. A number of convolutional neural network (CNN) architectures are used to detect and classify disease symptoms through the use of paddy plant image. The key metrics are used to assess the performance of the DL models. Computerized experiments indicate that the technology of DL based can be successfully used to identify and recognize paddy plant diseases with a high degree of accuracy. The DenseNet201 architectures are the best performing models in relation to classification accuracy and robustness as compared to the VGG16 and ResNet152V2 DL models. Developing a common comparative evaluation scheme of different CNN architectures in the same experimental conditions. The results identify the possibility of using DL techniques in creating automated systems of crop disease detection that can help farmers and agricultural specialists to diagnose and intervene in time.