A Hybrid Intelligent Framework Based on Deep Learning and Adaptive Fuzzy Diagnosis for Rice Leaf Disease Detection and Classification

Rising incidences of leaf diseases in rice for a more stable food security has become a global challenge. Therefore, there is a need to come up with more scientific diagnostic approaches. Most of the conventional means of disease recognition depend on visual inspection and interpretation by expert predicting the disease, a process that is tedious and subject to human bias. To address these issues, new approaches such as deep learning are very effective in overcoming taken up by using the imagination of a machine. This mechanical approach and the development of new algorithms based on CNNs and all-stage datasets of healthy and infected rice leaves could accelerate the detection process and enable people to act quickly. This article discusses deep learning methods specifically designed for diagnosing diseases on rice leaves. This is achieved by reviewing the entire process, from the conception to its full implementation in agriculture. The synthetic-data study reported used 400 images from four classes, split into 300 training images and 100 test images. The CNN achieved 86.00% accuracy and a macro F1-score of 85.65% after 12 training epochs. The adaptive fuzzy part reached 92.00% accuracy on a separate task of simulated severity classification.