Cultivating Health: Automated Sugarcane Disease Detection Through Deep Learning

Widely grown in tropical and subtropical climates, sugarcane is a significant commodity providing a major source of sugar, biofuels, and many by-products. But it is quite vulnerable to diseases such Mosaic, Red Rot, Rust, and Yellow Leaf Disease, which seriously lower crop output and quality and so jeopardize farmers' lives as well as the agricultural economy. Conventional methods of disease detection rely on time-consuming, labour-intensive manual inspection prone to mistake. This paper addresses leveraging deep learning for automated sugarcane disease diagnosis in order to solve this problem using the MobileNetV2 architecture, a lightweight convolutional neural network developed for real-time applications. Five classes-Healthy, Mosaic, Red Rot, Rust, and Yellow Leaf Disease-were created from 2,521 images of sugarcane leaves. Data preparation with resizing, standardizing, and augmentation was done to increase model generalization. Proposed is transfer learning using MobileNetV2 pre-trained on ImageNet and refined for the specific aim of disease classification. The model achieved 87% test set accuracy. This work emphasizes MobileNetV2's possibilities for real-time disease monitoring in the field and shows how well it detects sugarcane diseases, so offering a scalable, accurate, and effective method of early disease detection and management.