Enhancing Sugarcane Disease Classification Using Transfer Learning with Convolutional Neural Networks

Abstract

The economic implications of sugarcane diseases on local farmers in India are significant and multifaceted, affecting not only their immediate yields but also their overall financial stability and livelihoods. About 70 percent of India's rural households still primarily depend on agriculture for their livelihood. As a cash crop, sugarcane holds a very important place in India's agrarian economy. India is not only the largest consumer of sugar but also its second-largest producer. Identifying the diseases in their initial stages helps not only the farmer but also reduces the burden on the country in many aspects. This paper discusses DenseNet, VGG, and ConvNeXt for classifying diseases in sugarcane plants, along with the detailed experimentation conducted. Based on evaluation metrics, ConvNeXt outperforms with 96% accuracy compared to DenseNet and VGG architectures on sugarcane disease detection.