Cassava Leaf Disease Recognition Using Convolutional Neural Networks

Authors:

Abstract
This study aims to investigate various kinds of convolutional neural networks to classify cassava leaf diseases. The objectives are that we want to survey the performance of current well-known CNN models including VGGs, ResNet, DenseNet, and Inception in term of classification performance and plan to extend the study of cassava disease recognition in the future. From our experiments, the best model is DenseNet121 with brightness augmentation, which achieves classification accuracy at 94.32% and F1-score at 92.13%.

This publication has 12 references indexed in Scilit: