Deep Learning for Smart Farming Using Efficient Net-Based Disease Detection in Crops

Crops are a major source of food worldwide, which gives people food security and economic assistance for their survival. Climate change which causes drastic changes in the environment along with the immunity lack in crops causes a substantial decrease in growth and yield. The detection of ailments in the early stages is essential for the prevention of crops like tomatoes and potatoes. The detection of diseases has become an important step towards treating the crops to increase the yield. The inadequate knowledge and unavailability of resources for small-scale farmers are the main reasons for late treatment or no treatment. The advancement in technology like Deep Learning (DL) can assist more when compared to traditional methods of detection and treatment. Each disease has some visible patterns on the leaves which can be leveraged to detect the disease with the help of Convolutional Neural Networks (CNNs) and Computer Vision. This paper presents a hybrid DL model that combines the strengths of EfficientNetB0 and transformer-inspired multi-head attention to identify six classes of tomato and potato leaf conditions, including healthy and diseased states.