Image-Based Crop Disease Detection Using Machine Learning

Crop disease detection is critical for agricultural productivity and global food security. Traditional methods rely on labour-intensive field surveys prone to human error. This paper presents an image-based crop disease detection system using a hybrid Convolutional Neural Network (CNN) augmented with pre-trained AlexNet weights. The system captures leaf images, applies preprocessing (resizing, normalization, augmentation), and classifies diseases with high accuracy. A Flask-based web application enables real-time prediction accessible to farmers via smartphone. Trained on 2,000 field-collected images across potato, pepper, and tomato crops, the proposed model achieves approximately 97% classification accuracy, outperforming standalone classifiers including SVM, Logistic Regression, Decision Tree, and Naïve Bayes