Cassava Disease Classification with Knowledge Distillation for use in Constrained Devices

In this paper, we present a method for improving the performance of a deep learning model for cassava leaf disease classification on constrained devices using knowledge distillation (KD). KD involves transferring knowledge from a deeper teacher model to a shallow student network, enabling us to reduce the weight of the model without sacrificing its performance. Our proposed approach is aimed at creating a cost-effective and mobile-optimized solution for cassava leaf disease classification, which is of particular importance given the high carbohydrate content and susceptibility to viral infections of cassava. Through the incorporation of KD, we demonstrate the potential for our method to enhance the performance of deep learning models in resource-constrained environments.