Hybrid Approach of Sensors and Deep Learning for Cotton Plant Disease Detection

Every country's economy is mostly dependent on agro-based products, and India is regarded as having a strong cotton agriculture sector. Cotton plants that have been infected by several diseases may suffer catastrophic losses. Nonetheless, smart plant disease diagnostic tools assist in spotting and identifying cotton plant disease in its early stages, but they are limited by a lack of access to qualified and appropriate professional support. The majority of studies are concentrated on deep learning (DL) techniques to identify plant illnesses using photos, which are insufficiently accurate for the diseases of cotton plants. An integrated hybrid technique using sensors and a DL model to find and identify illnesses in the cotton plant has been developed to address these problems. Via several metrics, the sensor assists in monitoring the soil's quality. The two distinct datasets are taken into account for sensors and DL prediction separately and trained using VGG16, Resnet50, EfficientNetB7, MobileNet-v2 and Resnet152v2. For this study, the best performance is choosen VGG16 with 0.9965 AUC.