Classification of Plant Species with Transfer Learning-Based Methods

EasyChair Preprint 8447

Abstract

Plant medicine plays an important role in the environment, industrial and ecological fields, especially in agriculture. In recent years, global warming and loss of biodiversity have been experienced as a result of urban development and environmental destruction. As a result of global warming, there are a wide variety of negative events, from the depletion of the ozone layer to the death of many plant species. Just as depletion of the ozone layer can cause people to see harmful sun rays directly, similarly, different damages occur with the extinction of plant species. For the stated reasons, nature and human life are directly related to each other. To preserve this relationship, it is necessary to protect plant species that are beneficial to humanity. The first step in conserving plant species beneficial to humans is recognizing plants.  There are thousands of plant species on earth, named or unnamed. Categorizing and classifying so many types of information also require a great deal of computing power. Classification can be made by using distinctive features of the plant, such as flower, bark, fruit, and seed, or by using leaf feature. There are feature extraction-based, sparse representation-based, and subspace learning-based methods to classify plants as leaf-based. However, in this study, deep learning-based methods, which provide a strong structure in classification tasks without applying any preprocessing, have been preferred recently. In this study, a deep learning model based on transfer learning is proposed based on the MobileNet V2 architecture. The proposed structure is compared with the base MobileNet V2 model with performance metrics.

Keyphrases: Artificial Intelligience, MobileNet-V2, Plant types, Transfer Learning, deep learning

@booklet{EasyChair:8447,
  author    = {Halit Çetiner},
  title     = {Classification of Plant Species with Transfer Learning-Based Methods},
  howpublished = {EasyChair Preprint 8447},
  year      = {EasyChair, 2022}}