Crop Prediction and Plant Disease Detection Using Machine Learning

The Plant diseases are among the primary causes of reduced crop yield and quality, affecting global food production and agricultural sustainability. Conventional disease detection methods, which rely on manual observation and expert consultation, are often time-consuming, error-prone, and inaccessible to small-scale farmers. With the advancements in deep learning, automated image-based plant disease detection has become a promising solution. This paper proposes a Convolutional Neural Network (CNN)-based approach, leveraging a custom CNN architecture and DenseNet121 for classifying plant leaf diseases. A dataset of over 11,000 leaf images was preprocessed and augmented to improve robustness against variations in lighting, background, and image quality. The models were trained and evaluated using metrics such as accuracy, precision, recall, and F1-score. Experimental results demonstrate that the custom CNN achieved superior validation accuracy and stability compared to DenseNet121. The system was further deployed as a web application using Flask, providing farmers with real-time disease predictions and visual explanations through Grad-CAM. The proposed framework shows potential for real-world agricultural adoption, enabling early detection, reducing yield loss, and supporting sustainable farming practices.