Detecting tomato leaf diseases with convolutional neural networks and image processing using a Sri Lankan tomato leaf dataset

Abstract

Crops like tomatoes are vital to farmers' livelihoods in Sri Lanka, where agriculture is a key economic pillar. But growing tomatoes comes with a lot of difficulties, not the least of which is the possibility of certain diseases that can destroy crops. The timely implementation of interventions and reduction of losses are contingent upon the early discovery of these disorders. Using convolutional neural networks (CNNs) and image processing techniques, this study offers a novel solution to this problem by detecting tomato leaf illnesses. One unique aspect of this study is the use of a custom dataset made up of photos of Sri Lankan tomato leaves from several farms in Embilipitiya, Suriyawewa an area noted for being susceptible to several tomato illnesses. The dataset includes a variety of disease categories that are common in the local agricultural setting, such as tomato early blight, tomato Septoria leaf spot, tomato curl, and tomato leaf minor. The quality of the dataset is improved using pre-processing methods including segmentation and picture enhancement. The dataset is then used to train a CNN architecture for the purpose of classifying diseases. The efficiency of the suggested method is demonstrated by the experimental findings, which show that it can accurately identify and classify tomato leaf diseases. The system that has been built provides an automated and effective tool for early disease diagnosis, which facilitates timely intervention and efficient management approaches. Utilising a localised dataset improves the system's resilience and adaptability, which makes it ideal for implementation in Sri Lankan tomato farms.