Ensemble Learning based Plant Leaf Disease Classification Considering Deep Convolutional Features from Pre-trained CNN

In this modern era of advanced technology based agriculture or farming, the automatic detection of plant leaf diseases have become an important task for monitoring plants' health. Moreover, the classification of plant leaf diseases are also closely related to people's health. In the recent developments of automatic detection of plant leaf disease, AI-based algorithms are widely used. Among the AI-based algorithms deep learning (DL) models such as convolutional neural networks (CNNs) are mostly considered due to their ability to extract deep convolution features automatically. However, the training of a CNN model is a time consuming task. Therefore, in this work pre-trained CNN model is used for deep convolutional feature extraction. Moreover, for the leaf disease classification task using deep convolutional features the ensemble of machine learning (ML) model is used as a light weight model over DL models in terms of reducing the training cost. The proposed ensemble approach is tested with kaggle plant disease dataset and the accuracy obtained is 95.8%.