Disease Detection in Plant Leaves Using Transfer Learning

The paper proposes an automated approach for plant disease detection using transfer learning techniques. The study focuses on three critical agricultural crops-tea, potato, and tomato known for their economic significance and susceptibility to a range of diseases. Leveraging deep learning and transfer learning, the proposed system integrates pre-trained models such as VGG19, ResNet152V2, and InceptionV3, fine-tuned to classify plant leaf diseases with high accuracy. The dataset used in this study includes tea leaves with seven distinct diseases, potato leaves with three classes of diseases, and tomato leaves with ten disease classes. Pre-processing was done on these datasets to improve model training and guarantee homogeneity. This included resizing and data augmentation. The best-performing model for each crop was identified by analyzing the transfer learning models using a variety of measures, such as validation accuracy, confusion matrix analysis, and F1-score. By comparing various transfer learning models, this study demonstrates that deep learning-based approaches offer superior performance in plant disease detection. The improved accuracy and robustness of these models provide a more efficient solution for early disease identification, allowing for timely treatment and reduced crop losses. The ultimate goal of this paper is to support agricultural professionals and farmers in early disease detection, facilitating precision farming practices and contributing to increased agricultural productivity and sustainability. This technology has the potential to make a significant impact on food security and promote a more sustainable approach to agriculture.