CoTE Net: Recognizing Overlapped Unhealthy Leaves Using Hybrid Deep Learning Model

Productivity in agriculture is essential to the expansion and advancement of the world economy. Crop diseases have a detrimental impact on a nation's financial resources and agricultural productivity. The world's food supply is at risk due to plant diseases. However, diagnosing plant diseases is still challenging and time-consuming. Deep learning techniques have recently been used to analyze digital photos in order to automate the diagnosis of plant diseases and assist non-experts in identifying ill plants. Several "Deep Learning (DL)" applications that are used to identify damaged plant leaves and try to increase the detection accuracy have addressed the problem of the models' enormous parameter size.This work has used a plant village dataset with 27 classes and 51,806 images. The recognition, of disease objects using a proposed technique a CoTE Net model rather than the overall part of the image. The proposed hybrid model has been compared with VGG 16, ResNet 152 V2, EfficientNet V2M, and NAS NetLarge model and classify the single and overlapped diseases of the object. Infections in bacterial and septoria spots on leaves were accurately generated by the proposed model with 99.23% accuracy, 98.98% recall, 97.99% F1 score, and 99.85% accuracy. The suggested hybrid approach performs better than comparable existing approaches in the identification and classification of multi-class plant overlapping leaf diseases.