Deep Neural Networks in the field of Machine Learning (ML) are broadly used for deep
learning. Among many of DNN structures, the Convolutional Neural Networks (CNN) are
currently the main tool used for the image analysis and classification problems. Deep neural
networks have been highly successful in image classification problems. In this paper, we
have shown the use of deep neural networks for plant disease detection, through image
classification. This study provides a transfer learning-based solution for detecting multiple
diseases in several plant varieties using simple leaf images of healthy and diseased plants
taken from PlantVillage dataset. We have addressed a multi-class classification problem in
which the models were trained, validated and tested using 11,333 images from 10 different
classes containing 2 crop species and 8 diseases. Six different CNN architectures VGG16,
InceptionV3, Xception, Resnet50, MobileNet, and DenseNet121 are compared. We found
that DenseNet121 achieves best accuracy of 95.48 on test data.
Plant Disease Detection in Imbalanced Datasets using Efficient Convolutional Neural Networks with Stepwise Transfer Learning
