Revolutionizing Tomato Farming in Zambia: AI Deep Learning as a Tool for Smart Agriculture

The integration of artificial intelligence (AI) in agriculture represents a pivotal advancement in modern farming practices. AI has introduced a transformative dimension to conventional agriculture, offering innovative solutions to the multifaceted challenges facing global food production. The pursuit of heightened profitability through increased productivity underscores the necessity for agricultural sustainability and efficiency. This imperative calls for a departure from conventional farming methods and urges transition toward data-driven, intelligent systems. CM farms located 35 km southeast of Lusaka central district of Zambia, a country in sub-Saharan region, currently faces a challenge of inefficient disease detection in their tomato field due to inconsistent visits of the agronomist to provide expert guidance. When available, the agronomist uses manual methods of disease detection of row-by-row field inspection, which is both labor-intensive and laborious. Consequently, a Convolutional Neural Network model to facilitate real-time tomato leaf disease detection was built. A Raspberry pi was fitted with a camera for capturing and sending of tomato leaf images to a web-based application which was built with a cron job for automated mail transmission to the user. Three deep learning models, namely VGG 16, ResNet 50, and GoogleNet V2 were trained using the plant village dataset which also incorporated diseased and non- diseased leaf images from CM farms. Of the three models, ResNet 50 had the highest accuracy level of 95%, compared to GoogleNet V2 and VGG 16 with accuracy of 85% and 63%, respectively. Subsequently, ResNet was adopted t integrated in the web application.