Mangrove Tree Recognition using Deep Learning

Abstract

This thesis titled “Mangrove Tree Recognition using Deep Learning” is a very important topic in not only Computer Science and Engineering but also Botany. Recognition of different mangrove trees with high accuracy provides a lot of knowledge to people. However, because of the complex background of mangrove tree, the similarity between the different species of mangrove tree, and the differences among the same species of mangrove tree, there are some challenges in the recognition of mangrove tree images. This mangrove tree recognition is mainly based on the three features: leaf, root and fruit, which requires people to select features for recognition, and the accuracy is not very high. In this project, based on Inception-v3 model of TensorFlow platform, we use the transfer learning technology to retrain the mangrove tree category datasets, which can greatly improve the accuracy of mangrove tree recognition. We have used Google’s Inception-v3 model trained on 3000 images covering 5 different categories. We retrained the Inception model to classify the mangrove tree images, using the Tensorflow Library and achieved an overall accuracy of 99% on the images.