Innovative Cassava Disease Prediction through Comparative Deep Learning Methods

In the Philippines, cassava—a key staple crop—faces many diseases that severely impact its growth and yield. Since cassava is essential for food security, early detection of these diseases is crucial to prevent their spread. Traditionally, disease identification has relied on expert insights and the experience of local farmers. This study introduces an automated system that leverages pre-trained neural network models and deep learning techniques, including convolutional neural networks, to interpret complex data for disease identification. Using this system, our neural networks successfully identified anthocyanin and red leaf symptoms in cassava plants. Specifically, we employed the EfficientNetB3 model to improve disease recognition, even when data quality is inconsistent. With 10,791,220 parameters, this model was trained using transfer learning and achieved an accuracy of 89.90%, significantly above the baseline rate of 50.46%. The neural network classifier proved highly effective, accurately identifying cassava leaf disease even in unbalanced datasets. Our findings have practical implications for farmers and agricultural scientists, aiming to boost crop yields. This automated disease detection approach enhances the efficiency of disease data collection, aiding efforts to protect crops. We also propose additional research to further refine plant disease detection methods.