ITM Web Conf.
Volume 40, 2021International Conference on Automation, Computing and Communication 2021 (ICACC-2021)
|Number of page(s)||6|
|Published online||09 August 2021|
A Hybrid Approach for Landmark Recognition using Deep Local Features and Residual Network-50
Computer Engineering Department, Dwarkadas J Sanghvi College og Engineering, Mumbai, India.
* Corresponding author: firstname.lastname@example.org
As smartphones and mobile data become universal in modern society, the opportunities to interact with the real world would grow tremendously. Latest Technologies such as Oculus Rift and Google Glass attempt to bridge the gap between the virtual and the material. With advancements in computing speed and image recognition, the idea of augmented reality (AR) becomes more tangible. However, the sheer complexity of image processing and feature recognition is an area of concern for AR. A successful AR system must distinguish among many landmarks and identify or classify the existence of new landmarks. AR algorithms naturally lend themselves to using deep learning because of the adaptability required to various factors. This paper aims to develop and refine a deep learning algorithm that can distinguish landmarks from images using a google landmark database of known landmarks. Instance-level recognition is universally used in areas of Landmark recognition and is also the upcoming research area. Instance-level recognition is the brain behind Landmark recognition. As in Landmarks, the goal is to seek an instance of a common group instead of a group, requiring new deep learning techniques. In this paper, three different VGG16, Inceptionv3, and ResNet50 models are trained using the transfer learning technique and a Pure Convolutional Neural Network (CNN) model is also trained from scratch. This paper proposes a modified version of the ResNet50 model to increase the accuracy and performance of the models used. The revised version of Resnet50 contains an additional Deep Local Features (DeLF) processing layer before generating the final output.
Key words: Landmark Recognition / DeLF / ResNet50 / Transfer learning / CNN / Inceptionv3 / VGG16 / Instance-Level Recognition
© The Authors, published by EDP Sciences, 2021
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.