Sugarcane Disease Diagnosis Utilizing a Combined Resnet-Inception Deep Learning Technique and a Web Application

Biotic factors such as plant diseases have often been a thorn in the progress of plant growth and crop production all over the world, thereby limiting the food humans consume. The best-coordinated branch of farming is the production of sugarcane. Due to these perfect growing conditions, it has become the favorite crop form among farmers. It is directly linked to the sugar industry and is notable enough to have an influence on particular countries such as Brazil, India, and China. Although it ranks twenty-fifth among the world’s most commercially grown crops, sugarcane has the most significant output value of all crops grown for commercial purposes. On the other hand, several diseases threaten agricultural productivity and produce quality. Some of them farmers can see with their naked eyes when they inspect leaves on the plant. But many infections remain undetected, and farmers are overwhelmed by gigantic losses. Consequently, it is essential to identify the kind of infestation in a bid to reduce loss. The current research work develops a novel deep-learning architecture that identifies unhealthy sugarcane plants based on their leaves, stems, colors, etc. The work presents three cases where hybrid feature extractors involve two features, namely LBP and GLCM, which extract texture and shape-based features. The fused model of ResNet-Inception performed better than other deep learning approaches, such as GoogLeNet and AdaBoost, in the study. Some of the diseases include Yellow leaf disease, Cercospora leaf spot, Rust spots, Helminthosporium leaf spot, and red rot. Specifically, the proposed model achieved 98% accuracy, 97% precision, 95% recall, and 96% F1 score.