Enhanced Plant Disease Detection in Precision Agriculture Using Autoencoder and Transfer Learning

Agricultural productivity faces significant threats from plant diseases, leading to substantial economic losses and worsening global food insecurity. These diseases can drastically reduce crop yields, adversely affecting farmers' livelihoods and disrupting food supply chains. In response, agrotechnology and precision farming have emerged as critical solutions, emphasizing the need for sustainable crop management through accurate and timely disease detection. Early intervention is essential to minimize disease spread and enhance understanding of crop health. This study presents a comprehensive approach to plant disease classification using a combination of autoencoders for dimensionality reduction and a fine-tuned ResNet-9 architecture. The proposed methodology leverages the autoencoder to effectively reduce the dimensionality of leaf imagery data, thereby simplifying the feature space and improving computational efficiency. The reduced features are then fed into a ResNet-9 model, which has been fine-tuned to achieve high classification accuracy for various plant diseases. Our model achieves a remarkable accuracy of 99.2%. The development of this robust tool aims to facilitate early disease detection and improve crop management, contributing significantly to the field of precision agriculture. By providing precise and timely diagnoses ofplant health, deep learning models hold the potential to revolutionize agricultural practices, thereby enhancing yield and sustainability in food production.