Kareem Yasser, Amr Mohamed, Ahmed Amr, Loay Yehia, Eng. Samira Refaat, Dr. Fatma Helmy
5 January 2022
The main idea of this project is to study the recognition of monuments, as tourists can recognize the monuments by scanning them. The data of the monument will be displayed to the tourist. This initiative aims to make it easier for travelers to understand the backstories of tourist attractions. Many historical sites are devoid of information regarding the monument’s rich past and the fights that led to its creation and preservation to the present day. This project proposes an approach to recognizing monuments using Advanced deep learning techniques. For the implementation, a deep learning technique will be implemented which is outpainting. Moreover, because of the diversity of animals remaining present in the museums, it is hard to recognize the kind of animal in the Egyptian museum through its bodies; an Augmented Reality layer will be implemented and added to display information about the recognized monuments and to visualize these animals.
1.1 Purpose of this document
The purpose of this paper is to lay forth the specifics of the document. In addition, the documentation serves as a guide for developers as well as a record of product approval for the required features. The software implementation will be described in this paper. Algorithms and approaches are covered in the software implementation. The recognition of Egyptian monuments aims to provide travelers with comprehensive details and stories behind each monument to assist them in comprehending the works of art they see. The app will feature data from Egyptian monuments, allowing users to recognize them by using the camera (live-streaming) or submitting a video.
1.2 Scope of this document
This paper examines systems that are comparable to Go Museum, depicts the overview, scope, context of Go Museum system design, and addresses the mobile application’s goals and possible users’ characteristics. This document also goes through Go Museum’s functional and non-functional requirements, design limits, data design, and the application’s fundamental class diagram in detail. Finally, this paper goes through the various operating situations and the application’s timeline.
1.3 System Overview
• Collecting the dataset:
A set of Egyptian monuments (statues and artifacts) photos are collected. Splitting the dataset Firstly, each class in the dataset has to be augmented to increase its size. The dataset has to be split into 70% training, 20% testing, and 10% validating. The training part is used to train the deep learning approach and the testing part is used for testing and calculating the system performance.
• Applying deep learning technique:
– Outpainting technique using GAN: We intend to apply GAN neural network in outpainting for monument detection since this approach is guaranteed to detect visible and occluded monuments. However, there are many hyperparameters to be selected carefully for better design, such as the number of epochs.
• Performance evaluation: The performance of the approach is evaluated by calculating the mean square error and discriminator’s output values.
1.4 System Scope
Dataset for the monuments in the Egyptian museum. As there is no dataset published for the Egyptian museum monuments, the dataset needed to be collected by capturing the status and animals out there. Then the dataset was augmented to enhance the dataset size and add variety to the images.
• Advanced deep learning technique will be implemented which is the Generative Adversarial Network (GAN) for the recognition of monuments.
• Adding the out-painting feature using Generative Adversarial Network (GAN) for partially covered monuments.
• Text to speech: For each monument, there is a history behind it. To display these stories to
the user, the Text to speech feature will be implemented. The user will have the option either
to read the data behind the monument or listen to it.
• Augmented Reality layer that will be implemented and added to the recognized monument and animals remaining.