Mathematical and Visual Comprehension of Convolutional Neural Network Model for Identifying Crop Diseases

Abstract

Promptly detecting Crop Diseases (CD) mitigates adverse effects on plants. Convolutional Neural Networks (CNNs), especially Deep Learning (DL), are extensively used in computer vision (CV) and recognize patterns. Scholars have suggested several DL models for the detection of CD. Nonetheless, DL models need a substantial quantity of parameters, resulting in extended training durations and posing challenges for implementation on compact devices. This work discusses the Mathematical and Visual Comprehension of the CNN Model for Identifying Crop Diseases (MVC-CNN-CD). The inputs for image processing have been employed to produce a new collection of images, including their gradient images. High-level semantic characteristics are recovered from the initial and rebuilt images using convolutional and depthwise differentiated Convolutional Layering (CL). The softmax algorithm is utilized for categorization. Mathematical concepts in computational complexity have been presented. The proposed framework's effectiveness is assessed and compared with equivalent efforts using a publicly accessible CD dataset.