Abstract
Efficient crop management and yield optimization rely on accurate identification of sugarcane diseases. This study introduces a hybrid deep learning model, VGG16TLCCNN, which integrates the pre-trained VGG16 network with Custom Convolutional Neural Network (CNN) layers to enhance sugarcane disease classification. The proposed model employs transferlearning to leverage VGG16’s robust feature extraction while fine-tuning custom layers tailored to the unique visual patterns of sugarcane diseases. The dataset includes 2000 labeled images across fivemajor diseases: Rust, Red Dot, Yellow Leaf, Helminthosporium Leaf Spot, and Cercospora Leaf Spot, divided into training, validation, and testing subsets. Experimental results demonstrate that the hybrid model improves accuracy by 10–15% compared to conventional CNN and standalone VGG16 architectures, achieving superior generalization and reduced overfitting. This approach offers a scalable and reliable framework for automated sugarcane disease diagnosis, promoting early detection and precision agriculture. Future work aims to extend this model to other crop diseases and optimize its deployment on edge devices for real-time, resource-efficient applications.
Keywords