Plant Disease Detection Using Custom Deep Learning Model

Plant diseases have been of great concern in the agricultural sector as they destroy crops and result in low yields, which is not ideal in a world where billions of people need it to survive. Plants can be saved from this kind of destruction if the disease is detected early on. One of the ways of detecting diseased plants is by looking at the leaves. Diseased leaves look different from healthy ones. It’s nearly impossible for humans to check for dying crops in vast fields, but the process can be automated through computer vision and deep neural networks for highly accurate detections. Our goal today is to create a custom, lightweight DNN model that would be fine-tuned for the specific task of classifying diseased plants by using images of their leaves. We will test three custom models. One of the custom models will use EfficientNet-B5 as the backbone and this will be used for comparison with the lightweight models that include MobileNet V2 and MobileNet V3 Large.