Machine Learning Based Plant Diseases Detection: A Review
Hilman F. Pardede, Endang Suryawati, Dikdik Krisnandi, R. Sandra Yuwana, Vicky Zilvan
Abstract
Currently, applying machine learning technologies for management and monitoring of agricultural products are gaining significant interests. One of them is for plant diseases detection. Plant diseases are major cause of crop losses. The existence of automatic plant diseases detection is essential to predict the plant diseases as early as possible, and hence, reducing the crop losses. In this paper, we presents a review of advancement of machine learning technologies for plant diseases detection. Various approaches have been proposed in the field. In this review, we group them into two: works that focus on finding good features for shallow machine learning architectures such as SVM, those that focus on applying deep architectures of machine learning such as deep convolutional neural networks (CNN). For the later, we observe that the works either applied CNN as classifier or as feature learning. Our survey shows that while (CNN), have become the lead technologies in the field, replacing shallow architectures like SVM, many challenges still remain. First is the issue of robustness against environmental conditions. Second in on how to deal large variety of data and diseases with limited number of data.