Early Detection of Potato Crop Diseases Using Depth Wise Separable Convolutional Neural Networks: A Deep Learning Approach

Abstract

The early detection of crop diseases is crucial for farmers to manage and mitigate the impact on crop yields effectively. Visual inspection-based approaches used in conventional disease identification are arbitrary, labour-intensive, and prone to mistakes. To address this, deep learning techniques, particularly Depth Wise Separable Convolutional Neural Networks (DSCNNs), have emerged as promising solutions for automating disease detection in agriculture. In this study, we examine how the DSCNN architecture's various layers impact crop disease learning and prediction. We create, train, and fine-tune a few DSCNN models using various layer arrangements. We evaluate the models using a comprehensive dataset of crop images, including diseased and healthy plants. The experimental findings provide insightful information about how different layers affect disease detection precision and computational effectiveness, providing direction for creating effective deep-learning techniques for tasks requiring crop disease detection.