Harnessing Thermal Imaging with Hybrid Deep Learning for Early Paddy Disease Detection

Paddy, or rice, is one of the most significant crops in India, forming the staple diet for a large part of the population. In the Guntur region, more than 6–8 diseases are associated with paddy crops, but four are more prevalent, such as sheath blight, leaf blast, bacterial leaf blight, and brown spot. Many researchers detected diseases by using learning models on RGB data. In this case, diseases are detected only after visible symptoms; due to this, 30–40 % yield is affected. Early disease detection is worked with the thermal images, which show more sensitivity to various parameters. Thermal imaging has emerged as a vital tool for early disease detection, providing a non-invasive and rapid method to identify plant stress indicators. Experimental results exhibit reduced false positives and negatives, emphasizing the methodology's accessibility and effectiveness in early disease detection. The hybrid model InceptionResnetV2 attains superior performance of 99.01 % compared to the Resnet50 (94%) and Inception V3 (96.34 %) due to combining both advantages of each model.