TRANSFER LEARNING FOR TOMATO LEAF DISEASE DETECTION USING CONVOLUTIONAL NEURAL NETWORKS ON MOBILE PLATFORMS

ABSTRACT

In this paper, with the use of transfer learning on mobile platforms, a new approach for Tomato Leaf Disease Detection (TLDD) was accomplished utilising convolutional neural networks (CNNs). Our main goal was to create a model that would enable real-time mobile diagnostics by quickly and correctly identifying diseases of tomato leaves. The base CNN architecture used was fine-tuned on a carefully curated dataset of healthy and diseased tomato leaf images. Experimental results demonstrate outstanding performance in disease detection. The precision score achieved an impressive 100%, the recall value attained 96%, signifying a high proportion of true positives being identified. At a remarkable 97.96%, the F1 Score—which balances recall and precision—reached its peak. The deployment of TLDD system on mobile devices promotes decentralized decision-making and enhances efficiency of disease management strategies. This paper introduces a successful application of digital image classification on a mobile application platform. Our approach contributes to improving agricultural practices and ensuring food security for a growing population.

Keywords: Android Mobile Application, CNN, Image Classification, ML, Tomato leaf Dataset, Tomato Diseases, Transfer Learning.