GSAtt-CMNetV3: Pepper Leaf Disease Classification Using Osprey Optimization

Nowadays, the demand for pepper keeps on increasing with the increase in human population. Accurate diagnosis, flawless identification, and early detection of the lesions will improve the income of farmers. At present, deep learning (DL) based techniques assist farmers in identifying plant diseases with low cost and minimal time complexity. Hence, this study proposes a novel optimized DL model for classifying the presence and absence of pepper leaf disease using an effective feature learning process. The proposed study undergoes four major stages namely Pre-processing, Segmentation, Feature extraction, and Classification. In the pre-processing stage, initially, the input images are resized and the Improved Contrast Limited Adaptive Histogram Equalization (ICLAHE) technique is introduced to enhance the quality of the pepper leaf images. Then, the Kernelized Gravity-based Density Clustering (KGDC) technique is conquered to segment the diseased portions from the leaf images. Finally, the Gated Self-Attentive Convoluted MobileNetV3 (GSAtt-CMNetV3) technique is proposed to extract the features and classify the pepper leaf disease accurately. Moreover, a novel osprey optimization algorithm (Os-OA) is introduced to tune the parameters of the proposed DL model for enhancing the classification performance. The proposed study is implemented via the Python platform, and a publicly available Plant-Village dataset is utilized for the simulation process. Accuracy, precision and recall values achieved by the proposed pepper leaf disease classification for training percent 80 is 97.87%, 96.87% and 97.08% respectively.