Visual recognition using deep learning (Landmark recognition using deep learning)

• Main Author: Low, Xian Jun
• Other Authors: Yap Kim Hui
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Electrical and electronic engineering; Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/139713
• Institution: Nanyang Technological University

Ever wondered what the specific name of the place on that trip to some country is 5 years ago? Or that instance where someone posted a picture of a place that looks stunning but have no idea where and what it was called? The solution presented in this project aims to use deep learning methods by training an image classifier to accomplish the task. The project starts with a study on the state-of-the-art solutions that are available for this problem of landmark recognition. This includes the solutions that were presented for the Google landmark recognition challenge for both 2018 and 2019.The proposed solution for this project would be to train different models of imageclassifiers as well as doing a study on ensemble modeling to hopefully improve the accuracy of the overall performance. The individual models will be trained on a subset of the dataset that was used for the Google landmark challenge called the Google landmark dataset. This report will show how the various models that are trained will perform individually and how the ensemble will be able to compare with the various individual models.The end of this report includes some discussions about the results that were achieved as well as recommendations on improvements to this project.