Plant Leaf Classification Using Fine-Tuned Transfer Learning

Authors

  • Ch. Lakshmi Narayana, Kondapalli Venkata Ramana

DOI:

https://doi.org/10.17762/msea.v71i4.1204

Abstract

Enhancing our understanding and knowledge of the plants around us is highly essential and plays a crucial role in the field of medicine, economy, and organic agriculture. As a result, researchers have done lots of research to protect the biodiversity of plant life on the planet. In general, research in this discipline focuses on identifying plant species based on leaf images. Deep learning algorithms have shown highly promising results and they are now extensively used in research in this area. In this paper, we build CNN and transfer learning models for identifying plant species. This type of network is prone to overfitting, but regularization techniques like batch-normalization and dropout can solve these problems. In addition, transfer learning and parameter fine-tuning performed well on the same dataset. The proposed approaches were tested on the Leafsnap dataset with 184 classes and achieved a peak training accuracy of 98.64 and a peak testing accuracy of 95.78 percent.

Downloads

Published

2022-11-12

How to Cite

Ch. Lakshmi Narayana, Kondapalli Venkata Ramana. (2022). Plant Leaf Classification Using Fine-Tuned Transfer Learning. Mathematical Statistician and Engineering Applications, 71(4), 6051–6070. https://doi.org/10.17762/msea.v71i4.1204

Issue

Section

Articles