Efficient Fine-Tuned Deep Learning ResNet Model for Tomato Leaf Disease Classification

Globally, vegetable and fruit plants play a vital role in sustaining the nutritional needs of 7.5 billion people, with India securing the position of the second-largest producer of tomatoes. However, the widespread use of chemical agents, such as fungicides and bactericides, aimed at combating plant diseases, is posing a significant threat to the agro-ecosystem. This over-reliance on chemicals adversely affects crop production, both quantitatively and qualitatively. To address this issue comprehensively, a holistic approach is essential; encompassing both agricultural and social dimensions. This paper proposes a solution in the form of a transfer learning-based model designed for early and reliable detection of tomato leaf diseases, offering a crucial tool for farmers. The study focuses on utilizing the ResNet-152 architecture, employing fine-tuning by freezing specific layers while allowing training for the remainder. The model is configured with the optimizer RMSprop and a learning rate of 0.001. Four transfer learning-based models, namely VGG-19, DenseNet-201, ResNet-152 and the proposed tuned model, are considered and trained and tested on a dataset comprising tomato leaves classify leaves into 10 categories, comprising nine disorder classes and one healthy class. In the comparative analysis, these models’ performance is evaluated using accuracy, recall, precision, and F1-Score. The results show that the tuned model has a maximum achievable accuracy of 99.00%, demonstrating its importance in transfer learning for precise and early detection of tomato leaf diseases. This research underscores the significance of adopting advanced technologies in agriculture to address the challenges posed by plant diseases.