Exploring Deep Learning Methods for Identifying Plant Diseases: An In-Depth Evaluation and Comparative Study

The agricultural sector faces significant challenges due to the adverse effects of plant diseases, posing threats to both productivity and global food security. This study delved into the realm of employing advanced deep learning techniques to accurately identify plant diseases. This research investigates the efficacy of four deep learning state of art methods as Convolutional Neural Network (CNN), ResNet34, EfficientNet, and MobileNet-in the domain of plant disease identification. Utilizing a dataset sourced from Kaggle comprising 87,572 RGB images, the study explores the training and testing/validation phases of each algorithm. Training results reveal varied performances, with EfficientN et exhibiting the highest training accuracy of 99.30%, followed closely by ResNet34 with 98.94%. In the testing/validation phase, EfficientN et continues to outperform other algorithms, achieving a validation accuracy of 98.50%, while MobileNet follows with 95.10%. These findings underscore EfficientN et's superior performance in accurately identifying plant diseases, offering valuable insights for future research and practical applications in agriculture.