Plant Disease Classification Using Deep Bilinear CNN

Plant diseases have become a major threat in farming and provision of food. Various plant affected the natural growth plants infected are leading factors for loss crop production. The manual detection identification require careful observative examination through expertise. To overcome testing procedures an automated can be implied which provides faster, scalable precisive solutions. In this research, contributions our work threefold. Firstly, bi-linear convolution neural network (Bi-CNNs) leaf disease classification is proposed. Secondly, we fine-tune VGG pruned ResNets utilize them as feature extractors connect to fully connected dense networks. hyperparameters tuned reach faster convergence obtain better generalization during stochastic optimization Bi-CNN(s). Finally, proposed model designed leverage scalability by implying Bi-CNN into real-world application release it open-source. on variant criteria ranging from 10% 50%. These models evaluated gold-standard measures. study performance, samples were expanded 5x (i.e., 50%) found that deviation accuracy was quite low (0.27%) resembles consistent ability. larger obtained score 94.98% 38 distinct classes.