A Comparative Study on Different Transfer Learning Approaches for Identification of Plant Diseases

Plant diseases present a substantial threat to global food security, as they have the potential to significantly reduce crop yields. Detecting these plant diseases early and managing them effectively is crucial for ensuring sustainable agriculture. In recent years, image analysis, particularly transfer learning approaches, has emerged as a promising solution for identifying plant diseases from images. This paper comprehensively analyzes six transfer learning models, including Xception, InceptionV3, DenseNet121, VGG16, ResNet50, and MobileNetV2 for identifying plant diseases. The primary objective of this study is to evaluate the accuracy, and loss function values, both before and after fine-tuning and performance parameters of these models. The results of this study demonstrate that fine-tuning significantly improves the performance of all models. The study shows that the Xception model (98.41% testing accuracy and 0.079 loss value) achieves state-of-the-art performance, outperforming the other transfer learning models in terms of accuracy, loss function values, and performance parameters. By analyzing and comparing the various models, this research provides a comprehensive package for researchers and practitioners seeking to develop accurate and efficient techniques for identifying plant diseases.