A Fuzzy-Siamese Framework With Triplet Loss For Early Tomato Leaf Disease Diagnosis

Tomato yields in India, once highly productive, have recently declined due to factors like seasonal changes, disorders, insects, diseases and pests, resulting in lower-quality harvests. Diseases such as yellow leaf curl virus and Late blight have particularly affected the quality and capacity of tomatoes. Timely and automatic recognition of these diseases is essential for effective control and to prevent crop loss. Traditional deep learning models for detecting tomato leaf diseases demand significant memory and storage due to their extensive parameters. Although lightweight networks help reduce these demands, they still struggle with performance issues and imbalanced data. This paper presents a novel, triplet loss fuzzy based Siamese framework that achieves 98.7% accuracy in identifying tomato leaf diseases. This system is both efficient and lightweight, facilitating early disease detection and prediction.