Leaf identification using learning machines for seedling distribution in Thailand

Abstract

Despite being an essential natural resource, the forest cover in Thailand has been continuously declining as a result of population growth and industrial expansion. In an attempt to reverse the situation, the Thai government has drawn up a plan with an aim to increase the forest area to 40% of the country, and part of the promotion plans is free seedling distribution.However, although the seedlings can be visually classified, it requires solid botanical expertise. In this research, we propose to use machine learning to classify seedlings by leaf images. Eight traditional learning machines and four deep learning models are investigated. An image dataset of eight highly distributed seedlings’ leaves is built from seedlings provided by the Royal Forest Department of Thailand. The results show that SVM and DenseNet201 perform best for traditional and deep learning machines.