Development of a Web and Mobile Applications-Based Cassava Disease Classification Interface Using Convolutional Neural Network

Abstract

Cassava is one of the six food items identified as a critical food product for Africa, owing to its importance to African farmers' lives and ability to alter African economies. However, Cassava plant diseases have affected the yield of farmers significantly which has led to a decline in the agricultural production of cassava. Therefore, the aim of this research work is to develop a web and mobile applications-based system that would be able to detect cassava diseases based on its leaf images. To achieve this aim, pre-trained Convolutional Neural Network (CNN) models were selected using their previous performance and the application of transfer learning technique, new models were developed to classify cassava diseases based on the dataset curated and pre-processed. The best three models were selected: MobileNetV2, VGG16 and ResNet50. After training, the accuracy for each model was: 98%, 92% and 75% for MobileNetV2, VGG16 and ResNet50 respectively. Following evaluation of performance, the model with the best accuracy (MobileNetV2) was deployed using a web application interface. After deploying as a web and mobile apps interface, it was further tested to see how it would perform on the field. This research work was found capable of aiding farmers in being able to timely detect the type of disease affecting their cassava plants and the correct treatment to utilize; this also contribute towards Sustainable development goals (SDG) 2 and 12.