Time Interiors

ML Based Application For Interior Designing, Allowing

Customers To Get Suggestions For Optimal Wall Fixing
Items, Book Order And Pay Online.

Submitted by:
Faizan Abbas 2020-CS-409
Aizaz Ahmad 2020-CS-423

Supervised by: Dr. Tayybah Kiren

Department of Computer Science

Rachna College of Engineering and Technology
(A Constituent College of UET Lahore)
i
 ii

Declaration
We solemnly declare that the work presented in this report has not been submitted to
obtain another degree or professional qualification.

Faizan Abbas Aizaz Ahmad

iii
 Acknowledgments

All praise be to Almighty ALLAH who is the creator and the sustainer of everything.
Peace and countless blessings of Almighty Allah be upon the Last Prophet Muhammad,
upon his family, upon his pious companions and upon their followers till the day of
resurrection.

First and foremost, we express our deep gratitude and thanks to Allah Almighty for His
blessings that allowed us to complete this research successfully. Without His guidance,
we would not have expected to finish this project.

We genuinely want to express our heartfelt appreciation to Dr.Tayybah Kiran. She

provided immense support in gathering information, collecting data, and guiding us
throughout the project. We extend our gratitude not only for her guidance and patience
during this endeavor but also for her mentorship in our previous encounters with the
research world.

We also hold great respect for all the professors specially Mam Amna and Mam Natasha,
who have supported and helped us to reach the position we are in today. Each of them
has generously given their time, energy, and expertise, and we are grateful for their
contributions.

We would like to acknowledge our parents and family members for their sincere encour-
agement and inspiration throughout our project, as they have been a driving force in
our lives. We owe everything to them.

We are specially thankful for the company ’Time Interiors’ which collaborated with us.

In addition, there have been several individuals who knowingly or unknowingly assisted
us in successfully completing this project. We will always cherish and fondly remember
the joyful memories of the golden years spent at the RCET (UET Lahore) with gratitude.

Faizan Abbas

Aizaz Ahmad

iv
We cordially dedicate this effort to our Parents , Family members,
Teachers, Security Guards and to all our well-wishers who
commends us during this journey, appreciate our struggle and
uplift our confidence to achieve our goal, build our character and
hunt our ambition.

v
Contents

Acknowledgments iv

List of Figures ix

List of Tables x

Abstract xi

1 Introduction 1
1.1 Project Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Goals and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Scope of Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Literature Survey 4
2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Existing Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.5 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Problem Statement 10
3.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Need of Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4 Requirement Analysis and Design 12

4.1 System Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.1.1 Requirements Gathering . . . . . . . . . . . . . . . . . . . . . . . . 12
4.1.2 User Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.1.3 Non-Functional Requirements . . . . . . . . . . . . . . . . . . . . . 12
4.1.4 System Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.1.5 Project Timeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 System Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2.1 Use Case Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2.2 Activity Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2.3 Sequence diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2.4 Database Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.2.5 System Architecture Diagram . . . . . . . . . . . . . . . . . . . . . 16

vi
Contents vii

5 Implementation 18
5.1 Tools and Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

6 Testing 24
6.1 Types of Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
6.2 Test case preparation: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
6.3 Test Cases and Defect Tracking: . . . . . . . . . . . . . . . . . . . . . . . 25
6.3.1 Test Case of Login Attempts . . . . . . . . . . . . . . . . . . . . . 25

7 Results and Discussion 29

7.1 Results of App. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.1.1 Login . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.1.2 Register . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.1.3 Home Page . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.1.4 Model Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.1.5 Product Description . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.1.6 Price Calculator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.1.7 Cart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.1.8 Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.1.9 Staff Profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.1.10 Admin Side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.2 Results for Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.2.1 Resnet.50: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.2.2 User Defined: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.2.3 MobileNetV2: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

8 Conclusions and Future Work 36

8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

A User Manual 37
A.0.1 A.1 End User Manual . . . . . . . . . . . . . . . . . . . . . . . . . 37
A.0.1.1 A.1.1 User Registration . . . . . . . . . . . . . . . . . . . 37
A.0.1.2 A.1.2 Login . . . . . . . . . . . . . . . . . . . . . . . . . . 37
A.0.1.3 A.1.3 Reset Password . . . . . . . . . . . . . . . . . . . . 37
A.0.1.4 A.1.4 Browse and Select Materials . . . . . . . . . . . . . 38
A.0.1.5 A.1.5 Upload Wall Image for Analysis . . . . . . . . . . . 38
A.0.1.6 A.1.6 Place an Order . . . . . . . . . . . . . . . . . . . . 38
A.0.1.7 A.1.7 Submit User Feedback . . . . . . . . . . . . . . . . 38
A.0.2 A.2 Admin Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
A.0.2.1 A.2.1 Admin Login . . . . . . . . . . . . . . . . . . . . . 39
A.0.2.2 A.2.2 Manage Users . . . . . . . . . . . . . . . . . . . . . 39
A.0.2.3 A.2.3 Manage Products . . . . . . . . . . . . . . . . . . . 39
A.0.2.4 A.2.4 View Orders . . . . . . . . . . . . . . . . . . . . . . 39
A.0.2.5 A.2.5 View User Feedback . . . . . . . . . . . . . . . . . 39
A.0.2.6 A.2.6 Add Workers . . . . . . . . . . . . . . . . . . . . . 39
A.0.2.7 A.2.7 Generate Reports . . . . . . . . . . . . . . . . . . . 40
Contents viii

B Technical Manual 41

References 43
List of Figures

4.1 Project Time Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4.2 Use Case Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.3 Activity Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.4 Admin Sequence diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.5 User Sequence diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.6 Database diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.7 System Architecture Diagram . . . . . . . . . . . . . . . . . . . . . . . . . 17

5.1 Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2 Tech Stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.3 Abstract Level System Diagram . . . . . . . . . . . . . . . . . . . . . . . . 21
5.4 Wall Seepage Detection Model Methodology Diagram . . . . . . . . . . . 22

6.1 Test Case 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

6.2 Test Case 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6.3 Test Case 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6.4 Test Case 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

7.1 Login . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.2 Registraion credentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.3 Home Page . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
7.4 Model Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.5 Product Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
7.6 Price Calculator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.7 Cart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7.8 Reviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.9 Staff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.10 Admin Side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
7.11 Admin Side . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.12 Resnet.50 Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7.13 User Defined Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7.14 MobileNetV2 Model Results . . . . . . . . . . . . . . . . . . . . . . . . . . 35

B.1 Install VSCode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

ix
List of Tables

2.1 Competitor Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

x
 Abstract
The mobile application is developed for the interior designing company named ’Time
Interiors’ that promises to disrupt the selection and ordering process of the materials like
wallpaper, wall paneling, and wooden flooring. The application is aimed at the common
problems of wall defects or leaks that often prompt users to make wrong selections based
on no technical advice. Using a machine learning model the app is able to check wall
conditions from pictures uploaded by users and recommend specific materials for the wall
depending the users’ needs. Project started with designing a User Interface (UI) design
in the Figma application and then developing the frontend with Flutter and the backend
with Firebase. A dataset of wall images was gathered and used to train the CNN model
that was improved over multiple iterations in an effort to achieve high levels of accuracy
in wall condition detection. This machine learning model is embedded in the app so that
every function of the app. Some of the key features of the app include a comprehensive
catalog of wall and floor covering items, calculation of the materials and cost based on
provided dimensions, ideas of new trends, and professional tips. It also allows users to
order and pay for products with simplicity and efficiency. Moreover, the app checks wall
condition using uploaded picture and recommends the relevant product e.g. wallpaper
for good wall conditions and wall panels for damaged or seepaged walls. The overall goal
of Time Interiors will be to enhance user experience and ensure better user satisfaction
with interior design provision. The app allows users to order the materials with ease
and get the best services in terms of providing solutions for the walls. Machine learning
for the wall condition analysis allows enhancing the decision-making process, increasing
customer satisfaction, and online orders for the company and thus increasing the revenue.
Chapter 1

Introduction

1.1 Project Overview

Time Interiors is the mobile application that is presented as a unique innovation that
will enhance the user experience by providing the opportunity to select and purchase
materials for interior design at home. They are used by clients in a range of roles seeking
home or office ornaments or construction personnel, interior decorators, hotels and other
facilities. This app meets the company’s primary concerns like some of the walls used
are broken or there are problems like seepage and this makes it more difficult to decide
which material to use. Specifically, the app contains an intelligent model that, after
the user uploads the images, assesses the wall conditions to offer the proper material
recommendations users need to make to improve their wall space. The project entailed
the following major steps: Firstly, designing the user interface (UI) by providing novices
with a simple and attractive interface on Figma. Next was the frontend that was created
out of Flutter – a flexible and mobile device compatible means. The backend was also
designed with the help of Firebase to ensure that all the data related to the users,
orders and payment information is well protected and easily accessible. A dataset of
wall images was gathered and used to train a model known as Convolutional Neural
Network (CNN), which possesses high mean accuracy of detecting wall conditions. This
specific machine learning model contributes to the app by providing the user with an
accurate set of materials that suits to their interests.

1.2 Goals and Objectives

The major function of ‘Time Interiors app’ is to make the experience of interior designing
smoother and comfortable for customers. Key objectives include:

• Providing a Wide Range of Interior Design Materials: This application

is covering a wide range of products and services related with interior designs,
particularly with focus on wallpaper, wall paneling and wooden flooring to respond
to its customer demand.

1
Chapter 1. Introduction 2

• Offering Pre-informed Material Suggestions: A convolutional neural net-

works that takes in images provided by the users to assess wall conditions and
recommend materials that meet their needs such as fixing wall seepage or wall
damage.

• Simplifying the Ordering and Payment Process: To allow ease in orders

placement and payments to be made within the app, it should be easy and effective.

• Enhancing Customer Satisfaction: Providing expert design recommendations

and guidance that enable users to transform the spaces to be beautiful and more
useful.

• Increasing Online Orders and Profitability: The app should also help to
improve the selection of customers and offer expert advice, which in turn will help
to increase the number of online orders and ensure the company’s growth in profit.

1.3 Contribution
Time Interiors has several impacts on the interior design fraternity through the adoption
of innovative technology and design skills. The app’s key contributions include:

• Reducing Inconvenience from Damaged Walls or Seepage: There are func-

tions that accurately recommend materials and also, provide an analysis of the
state of the wall to suggest the required material for treatment of certain prob-
lems.

• Simplifying the Material Selection Process: Thus, the app eliminates the
numerous decisions making process and offers a convenient interface that helps to
avoid the instances of coming up with unique interior design materials.

• Leveraging Machine Learning for Data-Driven Recommendations: The

CNN model helps in the assessment of wall conditions based on images to generate
recommendations that would be of immense benefit to users; they would enable
them make proper decisions when it comes to the choice of materials to be used
in their walls.

• Facilitating Online Transactions: The app enables the users of the company to
order and pay for materials which will compel more people to buy online resulting
to more revenue.

• Combining Aesthetic Design with Practical Functionality:The app is not

only useful when it comes to selecting the beautiful materials but also perfect for
choosing the right material that will meet the requirements of the user’s room’s
conditions.
Chapter 1. Introduction 3

1.4 Scope of Project

The scope of the Time Interiors project covers the development and integration of various
components to create a user-friendly application. This includes:

• UI Design in Figma: Designing a user interface that would meet the users’
expectations while providing ease of use when navigating the interface.

• Frontend Development with Flutter: Designing the layout of the program

flows and transitions, and adapting the interface to different operating systems,
whether it is iOS or Android.

• Backend Implementation with Firebase: Applying Firebase in the backend

to store user data, orders, and payments, and, guaranteeing proper data storage
and security.

• Training and Integrating a CNN Model: Collecting a set of images of walls

and tuning the CNN model to give accurate results regarding wall conditions as
well as offering the proper material recommendations.

• Extensive Testing:Performing all the necessary tests to check that all features
and performances in the app function properly, and the user interface meets a high
standard, any problems being identified during the developmental and implement-
ing process.
Chapter 2

Literature Survey

2.1 Background
The industry of interior design has changed as a result of the use of technology. Before,
choosing and buying interior design materials required moving from one showroom to
another or browsing through magazines and catalogs. Nevertheless, it is necessary to
mention that the use of applications for portable devices and online services has changed
this process. These digital tools allow a user to present a variety of material and get
a glimpse of how those could look like in their environments. As the field progresses
there remains a focus on the continued polishing of the digital front end and greater
utilization of recommendation algorithms.

2.2 Related Work

Several researches carried out on the effectiveness of using technology in enhancing the
interior design by overcoming the challenges of material selection and ordering. So-
cial networking and buying apps, virtual interiors, online exhibitions, and idea rooms
developed for users have been created to offer virtual catalogs, interactive room configu-
rators, and online purchasing capabilities. Furthermore, actions like image recognition,
formulation of data analysis and machine learning have been used to personalize accom-
modation recommendations based on preference of the user and spatial needs. These
activities also cover the objectives of providing users with essential tools and information
for decision making in interiors design solutions.

2.3 Existing Systems

There are some similar systems at the present time, which are also as follows: Time
Interiors is similar in the way that it allows users search and order materials for inte-
rior designing. These include comprehensive database of materials, virtual rooms and
showing and options such as online ordering of materials. However, most of the existing
auto design systems have limitations in terms of identifying wall conditions and recom-
mending appropriate material types. This deficiency may lead to less relevant materials

4
Chapter 2. Literature Survey 5

for the needs of a particular user or room, especially when there are a large number of
options available.Given below few technical competitors of Time Interiors app so that
we can position ourselves in accordance to the features enlisted below. While these apps
are generally based on virtual environments and designing there is no information about
machine learning model of walls that is being developed in our app.

Competitors Time Interi- 1 Houzz 2 Room 3 Home 4 Magic

Features ors Planner Styler Plan

Analyzing Yes No No No No
Wall Condi-
tion
Comprehensive Yes Yes No No No
product list
Suggestion Yes No No Yes No
for Suitable
items
Online Buy- Yes Yes No Yes No
ing & Pur-
chasing
3D Visualiza- No Yes Yes Yes Yes
tion
Table 2.1: Competitor Analysis

2.4 Limitations
Even though there are significant improvements in the technological systems in the inte-
rior design industry, there are still some gaps that are observed in the current systems.
One type of weakness is the imprecision, which often refers to the case when the systems
offer quite general suggestions without taking into consideration personal peculiarities
of people. Furthermore, it can be noted that the vast majority of systems do not in-
clude certain parameters of the machine learning approach to estimate the condition of
walls and offer suitable materials. This can creep in and mean that users end up with
suboptimal material to work with, and dissatisfaction creeps in. In addition, some sys-
tems have involved ordering procedures that can be irritating to the user as well as the
credibility of what ought to be a smooth purchasing procedure. In order to enhance the
general perception of the sites and ensure that these platforms are optimally effective in
satisfying the growing demand in professional interior designing services, it is crucial to
consider these points out in an effort to come up with appropriate solutions.
Chapter 2. Literature Survey 6

2.5 Literature Review

In supervised learning, which involves CNNs, Cumbajin et al. [1] methodically reviewed
works on surface defect detection. conducted an analysis of the articles published be-
tween 2010 and 2021 and as a result, screened 59 primary studies. It was based on the
CNN for the surface’s characteristic detection in cascade with the detection of surface-
defect types through categorization of surfaces by distinguishable types that encompass
metal, buildings, ceramics and woods, special materials, and so on. The findings showed
that research on the metallic surface was the most common (62.71 per. ), and issues
in classification were the most prevalent (49.15 per. ). Pre-trained models, or transfer
learning, were used in 83. 05 per. of the studies, and data augmentation in 59. In total,
36 studies used data augmentation, while 23 studies used data augmentation with the
addition of other techniques. 32 per. The issues were also identified in terms of access to
details regarding cameras, lighting and image dimensions that were utilized in about half
of the editorial studies. ABC et al. also have pointed out for CNN as a more frequent
applying method for surface defect detection with a special reference toward the metal
industry and they expected more studies in the future regarding the joint application
of different problems in enhancing the accuracy. The review further emphasized the
need to confront the lighting issue since the presence of natural light in real industries
elaborated the image quality towards providing defect detection purposes.
From literature, Bhatt et al. [2] attempted to establish an innovative approach concern-
ing the detection of surface defects in manufacturing applications. He also pointed out
an important limitation of using image processing to solve problems of noise, changes
in light conditions and complex backgrounds. They underscored the increasing role of
deep learning in automating defect detection and presented a survey article that classi-
fied literature efforts in three ways: comparing with defect detection context, learning
techniques, and defects localization and classification strategies. To classify the existing
literature, the article proposed that future work could buckle down and discover deep
learning from those viewpoints such as the problem scope, machine learning form, and
system design form. The authors BCD et al. have noted increased usage of deep learning
in the field and proposed areas for future investigations.
To automatically detect and classify six typical sewer defects such as cracks, disjoints,
obstacles, residences, trees, and normal ones from multisource CCTV images, Zhou et al.
[3] proposed convolutional neural networks (CNNs)-based model. By using ways such as
data augmentation to increase model performance, they got average prediction accuracy
of 90 per. Subsequently, it was evident that the proposed method remains effective in
identifying sewer defects even if no prior knowledge is possessed by the engineer. In-
terestingly, the resul ting CNN model outperformed a state-of-the-art solution that the
authors proposed by retraining SqueezeNet using defect images while saving computa-
tional resources. This study demonstrated the prospective of deep learning technology
in accomplishing sewer defect classification and contributed to enhancing the knowledge
in the research area.
Chapter 2. Literature Survey 7

Consequently, Cha et al. [4] sought to address problems associated with identification
of civil infrastructure defects through incorporating a CNN based approach in visioning.
Unlike other conventional methods of image processing, this method does not involve a
priori analysis of features of defects such as the cracks on concrete and steel plans. The
CNN, trained with 40K Industrial images of 256 × 256 pixel resolution, made approxi-
mately 98 per cent accuracy. accuracy. The combined method with the help of sliding
window proved to work effectively even for a different structure and lighting conditions
when 55 out of the tested 1000 images of the size 5888×3564 were analyzed. Compar-
isons with the traditional techniques of edge detection showed that the proposed CNN
performed highly efficiently in the detection of concrete cracks in a real environment with
rigorous features such as pointer light source, shadow effect, and very narrow cracks.
In their empirical research, Butcher et al.[5], discussed the need for non-destructive test-
ing for defect identification in reinforced concrete by using ElectroMagnetic Anomaly
Detection (EMAD) technique. The article aimed at studying the novel application of
two neural network methodologies, ESNs and ELMs for automating the analysis of
electromagnetic signals that pertains to the reinforcement steel. For a one-tendon rein-
forced structure, LELMs achieved a higher performance compared to classical ELMs; for
two ESN architectures, the highest performance was obtained for the mesh-reinforced
concrete structure. The study was intended to optimize the time-costly data analysis
process to improve it with the faster and more efficient training offered by these neural
network approaches compared with others.
Similarly, Wu et al. [6] admitted that the RF-FEM in geotechnical engineering was
largely problematic from the computational perspective, and thus, they claimed that
to tackle this problem, they designed a deep learning model called the CNN surrogate
model that adopted the original VGG6 perspective. The output from the random field
method described above was in tabular form and was preprocessed into an image for-
mat as shown in the figure after which the CNN surrogate model replaced the RF-FEM
simulations. When used to evaluate wall deflection in the braced excavation taking into
account the spatial variability The results derived from the application of this method
were positively compared to those obtained employing the FEM The results from this
study using VGG6 indicate promising future for the reliability analysis of the spatial
variability of the soils for the model. The study also discussed the XGBoost model for
the purpose of comparison and the performance of the Hybrid model of fusion of both
the CNN with XGBoost was explored and stressed on the effectiveness and transferabil-
ity aspects of the proposed CNN model.
Chua et al.[7] in their study only used a neural network algorithm known as hybrid evo-
lutionary Bayesian back-propagation (EBBP) neural network for modeling SSI behavior
of deep excavations in clays.Chua et al.[7] Neural network in EBBP have used genetic al-
gorithms with the gradient descent method within a Bayesian framework in order to set
parameters for learning the problem; these steps are important to deal with the learning
complexity and to assess the uncertainty of statistical data. The analysis involved a
Chapter 2. Literature Survey 8

large Performance Database of braced excavation, data obtained through finite element
analysis. Preliminary parametric studies showed coherent and comprehensive patterns,
and the comparisons with back-analysis of instrumented data from the literature also
gave fairly good predictions with measured values. The trained neural network model is
presented to be a simple and viable method in giving maximum wall deflection predic-
tion during the preliminary design of braced excavations in clay, possible inputs such as
wall stiffness, support stiffness, stress state, non-uniform ground conditions, and depth
dependency of the soil properties, and bar inside the plots of results for error estimation.
For instance in their paper on generic cost effective Automated Surface Inspection (ASI)
that requires little to no training data Ren et al [8]. The method comprises of extracting
features from image patches by using a transfer learning based pretrained deep learning
network then using classifier trained from the images patch features and applying it pixel
wise to the input image by convolving the trained classifier with the input image. Evalu-
ation on three benchmark datasets and one industry dataset for image classification and
defect segmentation evidenced the efficiency of the proposed algorithm. The proposed
method was found to be more accurate than current benchmarks and further achieved
enhancements of accuracy over benchmarks ranging from 0. 66 per. to 25. 50 per. has
tried in the field of image classification and able to minimize the error escape rates to
a greater extent by 6. 00 per. to 19. 00 per. in defect segmentation. Furthermore, the
effectiveness of the approach was modeled with achieving 0. 0 per.
In their study Chatzhi et al.[9] incorporated new approaches in teh finite element for-
mulation(XFEM) with Genetic Algorithms (GAs) for optimum flaw identification in
structure. A forward problem was simulated by using XFEM and a new genetic algo-
rithm was presented to accelerate the convergence, reduce the chances of getting trapped
at local optima and also to obtain an error value comparing the sensor readings and data
obtained from solving the forward problem. In the same generality, using an XFEM for-
mulation of an ellipse shape for manufacturing a hole helps to identify different types of
flaws and forms. The current investigation involved the experimental validation of the
approach for a general 2D plate having an arbitrary crack and indicated the efficacy of
the proposed XFEM-GA algorithm for the identification of general flaws through the
application of convergence analysis for benchmark problems and experimental analysis.
In their work, the authors Wang et al. [10] identified the problem of having fast and ef-
ficient AQVI on the current generation of products. In contrast to the hand-engineered
feature-based detection method used in the previous experiments, the authors imple-
mented a new deep convolutional neural network (CNN) that could learn the powerful
features for the defect detection on its own. CNNs proved efficiency in the current ex-
periment in detecting the defects with minimal reliance on the prior information about
the images, besides being able to perform well even in the presence of noise. The eval-
uation of the CNN model based on a benchmark dataset revealed that the proposed
technique was highly effective in detecting defects rapidly and yielded better results
than state-of-art methods in the product quality monitoring field.
Chapter 2. Literature Survey 9

Authors Dataset Methodology Results Classes App

Cumbajin Not Men- CNN models Not Men- Not Men- No
et al.[1] tioned with supervised tioned tioned
learning
Bhatt et Not Men- defect localiza- Not Men- 38 Classes No
al.[2] tioned tion and clas- tioned
sification meth-
ods
Zhou et Random CNN models 95% Accu- 2 No
al.[3] Images racy
Cha et Not Men- CNN model 98% accu- 2 Classes No
al.[4] tioned racy
Butcher Not Men- ElectroMagnetic Not Men- Not Men- No
et al.[5] tioned Anomaly tioned tioned
Detection
(EMAD) tech-
nique
Wu et Not Men- CNN) surro- Not Men- Not Men- No
al.[6] tioned gate model tioned tioned
based on
classical archi-
tecture, VGG6,
to perform
random field
finite element
analyses
Chua et Not Men- hybrid evo- Not Men- Not Men- No
al.[7] tioned lutionary tioned tioned
Bayesian back-
propagation
(EBBP) neu-
ral network
algorithm
Ren et Not Men- Automated 97% Seven No
al.[8] tioned surface inspec-
tion (ASI)
Chatzi et Not Men- element formu- Not Men- 3 Classes No
al.[9] tioned lation (XFEM) tioned
combined with
genetic algo-
rithms (GAs)
Wang et Not Men- CNN Not Men- Not Men- No
al.[10] tioned tioned tioned
Our Real Ima- MobileNetV2 Accuracy Two Yes
Study gies 98%, Val-
accuracy
94%, Preci-
sion 94%
Table 2.2: Literature Review
Chapter 3

Problem Statement

Interior designing is a profession that greatly impacts the quality of our spaces in and
around the home or the workplace by not only the appearance but also the usability and
ergonomics of the spaces. Even in the present age with all available modern amenities,
interior design and decorating has one major problem – the frustration that comes with
wall cracks and seepage problems. These problems raise customers’ improper selection
of materials, which can be attributed to the absence of guidance and experience. For
instance, a customer might decide on a wallpaper that is good looking yet has a problem
of being installed on a wall that has seepage problems and the problem was made worse
thus causing the customer to be dissatisfied. However, most firms in the interior design
space find it difficult to get orders through the Internet because the processes involved
in the selection of the materials can be time consuming and cumbersome to many of the
buyers. This implies that there is a big percentage of customers ordering products that
are unique these are activities that require a lot of time and manpower from the side of
the company as well as the customer. Now, there is a clear requirement for an integrated
solution that provides guidance for selecting materials, organizes an online store, and
increases customer satisfaction, making companies more effective and profitable in their
production.

3.1 Motivation
The essence of the Time Interiors project lies in the attempt to respond to these ma-
jor issues in the context of interior design. The design target for this project is to
apply technology to improve the experience of users and consequently also of interior
design businesses. Mobile applications and machine learning offer a unique possibil-
ity to revolutionize the choice and purchase of interiors materials by customers. Thus,
by implementing an embedded Convolutional Neural Network (CNN) model capable to
determine wall conditions from images, it is possible to offer accurate material require-
ments for every customer. Personalization is not only unique but critical in this case to
help customers make good decisions and avoid selecting wrong materials. Therefore, by
creating a mobile application which will incorporate this model, we are likely to make
10
Chapter 3. Problem Statement 11

it easy for users to order for materials and make payments from the comfort of their
homes. This project is motivated by the vision of changing the state of play in the
interior design market making it more effective, efficient and easy to use.

3.2 Need of Project

There are several important reasons for the necessity of the Time Interiors project that
can be described as essential problems for both customers and companies focused on
interior design. First, clients are challenged with the question of which materials to
use when undertaking interior modifications; this is especially so in cases such as wall
leakage or damage. They may end up making a decision that will not be in their best
interest and complications and general dissatisfaction may arise. This need is met in this
project by giving material suggestions after studying the wall conditions to ensure that
customers make right choices on materials that they require for the wall which would be
appealing to the eyes and appropriate for the wall. Secondly, the current trend where
interior design firms mostly worked on price quotations from clients is also a cause of
concern. Made to order is rather a lengthy process, which entails a lot of effort from the
buyer, as well as from the seller. This project is developed to enhance the current system
where customers needs to place a specific order for a particular material by simplifying
the selection, ordering and payment process and offering it directly to the customers.
By expounding on these processes, we can increase the levels of customer satisfaction,
customer orders through the internet leading to higher profits for firms in interior design.
Furthermore,, application of sophisticated tools and techniques for decision making like,
machine learning, mobile application development, is mandatory for surviving the cut
throat competition of the modern era. To fulfil this need, the Time Interiors model
developed in this paper integrates a CNN model with a user-friendly mobile application,
which provides a one-stop solution for the existing and subscribed customers as well as
the potential customers. Therefore, this project is needed in the interior design field to
fill the existing gap in translating technology into tangible application, while contributing
to positive development in interior design.
Chapter 4

Requirement Analysis and Design

4.1 System Analysis

4.1.1 Requirements Gathering
The requirements gathering process is one of the essential steps in the system devel-
opment process, so the goal is to define the needs of individuals interested in the final
product. This would include the consultation of people involved in the Time Interiors
project including interior designers, potential users of the application as well as technical
gurus in coming up with the major functionality/ feature requirements of the applica-
tion. Such surveys as questionnaires, interviews with users, and studies of other interior
design applications were conducted to identify its problem areas and untapped potential.
It proved to be effective in making sure that all needs to be met were met to enable a
good foundation that would be expanded in the subsequent phases of development to
be met.

4.1.2 User Requirements

User requirements are focused on matching end-users’ needs by the system that is devel-
oped so that they can interact with it and find it efficient, usable and easy to operate.
In our case, for example, buyers and sellers on Time Interiors require the application
to allow users to access various interior designs, including wallpapers, paneling, and
wooden flooring. They also need the application to be able to diagnose from the images
of walls uploaded their conditions and recommend the suitable material. Besides, users
require a convenient ordering system and simple way to pay for the product, popular
design trends, and recommendations from experts. These requirements were identified
by conducting several discussions with users of the application and market research in
order to select the actual client requirements.

4.1.3 Non-Functional Requirements

Non-functional requirements describe necessary and sufficient conditions in which the
system has to work. For this project, these are the performance features, usability,

12
Chapter 4. Requirement Analysis and Design 13

reliability and security. Ideally, the app must be optimally responsive and the user-
interface smooth on both the iOS and Android operating systems. It should be able to
respond to multiple users at the same time and protect users’ data, including personal
information and credit card numbers. Also, the app must be easy to maintain and
extend to include new content and other functions in the future.

4.1.4 System Requirements

System requirements are the requirements for the hardware and software necessary to
both build and support the application. Front-end of the application is developed by
using Flutter, therefore appropriate development environment and devices to test on
both iOS and Android platforms must be present. The back-end is implemented using
Firebase and this requires Firebase services to support data operation, authentication
and cloud storage. Moreover, the machine learning model for wall condition analysis is
proposed using TensorFlow, thus the model needs specific libraries and computational
resources in order to train and test the model.

4.1.5 Project Timeline

The project timeline of ’Time Interiors’ is a flowing and sequenced plan of activities
showing phases and milestones of the project. They consist of the collection of require-
ments, system design, development, testing, and deployment. Each phase is given a
certain number of days with start and end dates so that the project plan runs in an
orderly manner and does not lag behind. Some of these are the UI design mark, inte-
gration of the machine learning model mark, and the final test mark to help in tracking
the progress and completion of the project.

Figure 4.1: Project Time Line

Chapter 4. Requirement Analysis and Design 14

4.2 System Design

4.2.1 Use Case Diagram
The use case diagram for the Time Interiors app illustrates the interactions between the
users and the system. It identifies the primary actors (users) and their interactions with
the system functionalities such as browsing materials, uploading wall images, receiving
material suggestions, placing orders, and making payments.

Figure 4.2: Use Case Diagram

4.2.2 Activity Diagram

The activity diagram outlines the workflow of the key processes in the app. It depicts the
sequence of activities from browsing materials to placing an order, highlighting decision
points and parallel processes.
Chapter 4. Requirement Analysis and Design 15

Figure 4.3: Activity Diagram

4.2.3 Sequence diagram

The sequence diagram shows the interaction between the user and the system over time.
It details how a user uploads a wall image, receives suggestions, and completes an order.

Figure 4.4: Admin Sequence diagram

Chapter 4. Requirement Analysis and Design 16

Figure 4.5: User Sequence diagram

4.2.4 Database Design

The database design includes tables for storing user information, products, orders, and
workers and review details. Each table is defined with its respective attributes and
relationships.

Figure 4.6: Database diagram

4.2.5 System Architecture Diagram

The system architecture diagram shows the relations between the systems and help in
the significant understanding of how various components relate within the systems. It
Chapter 4. Requirement Analysis and Design 17

comprises of the user interface, the back-end services for data handling, the machine
learning model and the database. Here is a diagram representing the interior of Time’s
system architecture.

Figure 4.7: System Architecture Diagram

Chapter 5

Implementation

The implementation phase of the Time Interiors project involves the planning, design,
and theoretical work are translated into a functional, user-friendly mobile application.
This phase includes the practical application of tools and technologies, that are grouped
with a methodology to ensure that the development process is efficient and the final
product meets the users needs. The following sections explain the specific tools and
technologies used, as well as the methodology employed to bring the Time Interiors app
to life.

5.1 Tools and Technologies

We utilized a suitable variety of tools and technologies to develop an efficient, scalable,
and user centric mobile application. Each tool was selected for its specific strengths
and compatibility with the project’s requirements, ensuring a consistent integration of
various components.

Visual Studio Code (VSCode): VSCode was used as the main code editor for the
work done during the project. Due to its lightness and supported powerful extensions
and customizing options, the framework is suitable for creating both the frontend and
backend of a website. Debugging tools, the in-built terminal, and the Git interface
helped the developers code, compile, and maintain the application easily.

Android Studio: Along with VSCode we used Android Studio as the editor, Android
Studio was instrumental in the development and testing of the android build of the
application. Indeed, the efficiency of the application under test was considered on various
versions of Android devices and operation systems thanks to emulator and device testing
of Android Studio. This feature, which offers great support for both Flutter and Dart,
was beneficial for achieving better performance on Android devices.

Google Colab: Google Colab was used when it comes to the development and / or
training of the machine learning model. This cloud-based concept enabled what was
needed for training of the Convolutional Neural Network (CNN) on the images of walls.

18
Chapter 5. Implementation 19

The support of Google Drive and native TensorFlow library integration of Google Colab
made it a suitable arena for testing and redesigning the machine learning model, in
terms of multiple iterations.

Flutter and Dart:Flutter development coupled with the Dart programming language
was selected for the frontend development of the Time Interiors mobile application. The
relative advantage of flutter in terms of using one code base to produce applications for
both the iOS and Android platforms minimized development time and energy. Flutter
was combined with Dart – a language that had clean and concise syntax with huge
potential in application development for building complex applications. This made the
user interface both, reactive and visually acceptable.

Figma:The design of the user interface (UI) of the Time Interiors application was done
using Figma. This collaborative design tool enabled the designers to design interac-
tive and mock-up designs within the application hence easily sought feedback from the
stakeholders and the prospective users. Being a cloud-based design tool, Figma had a
collaborative interface that would allow the various teams to make modifications to the
designs and to review the changes immediately.

Figure 5.1: Tools

TensorFlow: Tensorflow was used to implement the CNN model which is responsible
for the wall image analysis of the images uploaded by the users. Before diving into the
solution, I would like to note that TensorFlow, the open-source library for machine learn-
ing and deep learning, has all the tools needed to build the model and train it efficiently.
TensorFlow Lite was then employed to fine-tune the model for mobile application, so
that it could actually run on users’ devices fast enough.

Firebase:Firebase was selected as the backend solution for its real-time database, au-
thentication, and storage. Firebase easily adapted with Flutter to provide synchronous
data retrieval, secure user identity, and user-submitted images management. Further-
more, the scalable nature of the Firebase platform guaranteed that the app could handle
the larger numbers of users and higher data volumes without compromising performance.
Chapter 5. Implementation 20

Figure 5.2: Tech Stack

5.2 Methodology
The specific mobile application called Time Interiors was created following an agile
framework for development. This was instrumental in ensuring that new changes could
easily be integrated into the existing project and at the same time the users’ input could
easily be incorporated into the project development processes.

Planning Phase: The project started with an identification and planning phase in
which the roles, responsibilities, goals, and strategies of the project were to be estab-
lished, critical path and potential risks were to be determined, and time frame for the
entire project was to be set. This phase involved activities such as surveying and in-
terviewing the interior designers, hosting focus group discussions, and maybe observing
the potential users. They also established a list of the analyzed requirements which was
prioritized and recorded for the use in the later stages of development.

Design Phase: After the planning phase came the design phase to create blueprints
and maps for implementing the plans. The design team of the appropriately named
Figma designed all the UI/UX interfaces and developed functional prototypes. These
designs were looped a number of times where stakholders were allowed to approve the
design and make modifications that would make the user interface more appealing to
the eye. Using these prototypes were beneficial in that it made it easier to establish an
overall vision for the app and allowed for evaluation of how usable the final design would
be before fully implementing it.
Chapter 5. Implementation 21

Figure 5.3: Abstract Level System Diagram

Development Phase: The four that were defined as the development phase were
further broken down into multiple sprints where each of them wasearing two weeks.
Throughout the sprints, the actual development strategy followed by the team was
about on setting goals and objectives for each period and adding relevant features and
functionalities accordingly. The web frontend of the application was built on Flutter
and Dart, backend services and database chosen was Firebase. The development of
the machine learning model aimed at being in parallel with the creation of the mobile
application through regular integration and testing of the Colab Google environment
and TensorFlow.

Model Training: The Wall Seepage Detection feature within this all is a key function
to help the users to detect the wall condition problems bit easily. The methodology for
implementing this feature involves the following steps:The methodology for implement-
ing this feature involves the following steps:

1. Data Collection: Gathered a diverse dataset by capturing different wall images by

ourselves and classify that images into seepage and clear categories.

2. Preprocessing: Resizing, scaling, zooming and batching is involved in the prepro-

cessing of the gathered images to get the better results.
Chapter 5. Implementation 22

3. Machine Learning Model Development: Utilize machine learning techniques, such

as convolutional neural networks (CNNs), to develop a robust model for wall seep-
age detection. Train the model using the preprocessed dataset, fine tuning pa-
rameters and architecture to maximize accuracy and reliability. Experiment with
four different CNN models: TFLite, ResNet50, MobileNetV2, and a User Defined
architecture.

4. Integration: Integrate the trained model into the Time interiors mobile applica-
tion, ensuring seamless functionality and user experience. Implement an intuitive
interface that allows users to capture and upload images of walls for analysis.

Figure 5.4: Wall Seepage Detection Model Methodology Diagram

Testing Phase:To underline, testing was an essential component of each examined

sprint, as it aimed to determine whether the delivered features were functioning properly
and fulfilled the corresponding requirements. The team also performed unit testing to
Chapter 5. Implementation 23

confirm proper operations of the components used in the development of the application,
integration testing to ascertain that various constitutions of the application were fully
operational, and user acceptance testing (UAT) that tested the total value proposition of
the solution. These modifications needed to be made to heed the feedback and correct
flaws before the final release, which is why getting user acceptance by the UAT was
important.

In conclusion, the Time Interiors app was implemented through a difficult and iterative
process that made use of modern tools and technologies in an agile framework. With
this method, the app was developed efficiently satisfied user needs, and could be easily
updated to meet new requirements.
Chapter 6

Testing

Testing is a critical step in the development process where it is ascertained whether

the application has the required qualities and works as expected within the anticipated
contexts. In this section, the kind of testing done, how testing scripts are developed,
and how defects are reported is presented.

6.1 Types of Testing

Unit Testing: This type of testing was centered on ascertaining the fitness of the par-
ticular segment or unit of the application. In this way, each function, method, or class
was tested individually to check that it worked as it was designed to work.

Integration Testing: Regulation tests were done in order to ensure that individual units
or components of the application interacted appropriately. This served to check that
interfaces between modules were working as was expected.

System Testing: This end-to-end testing type tested the entire application in a complete
environment, thus being very useful. It made sure that all the components of the entire
system would conform with the established requirements.

Acceptance Testing (UAT): To ensure that the application complied with the needs and
expectation of the actual users, the UAT was performed with real users. This type of
testing was useful in exposing any problems that could be related to usability or the
absence of some functionality.

Regression Testing: Whenever additions or modifications were made to the code base,
the regression tests checked whether the functionalities relevant before the change were
compromised in anyway. It was vital in ensuring that the application was stable and
sustainable in the long run.

6.2 Test case preparation:

During test case preparation, specific business cases were created to use on the applica-
tion being tested. The general format for each test case was the steps followed in this

24
Chapter 6. Testing 25

test, the result that should be obtained, conditions or prerequisites before running this
test and conditions after running this test. The test cases were then elaborated to cover
both usual and corner cases so that no possible outcome is left unconsidered.

6.3 Test Cases and Defect Tracking:

Defect tracking was done with the help of a defect tracking system to compare the actual
defects found in the testing phase. Every single defect was given an ID and information
on the test case number it is linked to, the type of issues, how it could be replicated, its
status etc.

6.3.1 Test Case of Login Attempts

This test case verifies the login functionality under various conditions to ensure that
only authorized users can access the system. Test Case 1: Valid Username and Valid
Passketøy
Inputs: It refers to the combination of a valid username with a valid password.
Expected Result: This means the user has logged in successfully, and has been
redirected to the home page.
Actual Result: The user is successfully logged in and redirect to the home page of
the application.
Defect Status: None.
Test Case 2: This accounts for the possibility of either enter an invalid username
or password
Inputs: Incorrect credentials, either the username or the password entered is
wrong.
Expected Result: It contains a message for unauthorized access, meaning the user
has entered wrong username or password.
Actual Result: This was the message that popped up on the screen: Login failure
: Invalid username or password.
Defect Status: None.
Test Case 4: Five Bad Logins
Inputs: Users with correct names, incorrect passwords entered five times.
Expected Result: Account locked for a short period when the account has been
entered wrong credentials five times, and the user is prompted to try after some
time.
Actual Result: Warning message to be shown after five times attempts made by
the user.
Chapter 6. Testing 26

Defect Status: None.

Because of the Flutter testing framework’s excellent integration with the Flutter
development environment, we chose it for our project because it was a very suitable
and effective fit. This framework allows for various test kinds, such as widget tests
for GUI components, integration tests for complete application flows, and unit
tests for individual functions. The main advantages of using Flutter’s testing
framework involve:
Test Case 1: Valid Username and Password
This test case verifies that a user can successfully log in with a valid username and
password. The expected result is that the user is navigated to the home screen
upon successful authentication.

Figure 6.1: Test Case 1

Test Case 2: Valid Item Details

This test case checks that valid item details are displayed correctly in the item
details screen. The expected result is that the correct item details are shown when
an item is selected.
Chapter 6. Testing 27

Figure 6.2: Test Case 2

Test Case 3: Valid Worker Profiles

This test case verifies that valid worker profiles can be viewed correctly. The ex-
pected result is that the worker profile screen displays accurate information about
the worker.

Figure 6.3: Test Case 3

Test Case 4: Valid Order Booking

This test case ensures that a user can successfully book an order with valid details.
The expected result is a confirmation message or screen indicating that the order
has been placed.
Chapter 6. Testing 28

Figure 6.4: Test Case 4

These test cases cover the basic functionalities of login attempts, item details
viewing, worker profile viewing, and order booking in the Time Interiors app.
Each test uses the Flutter testing framework to simulate user interactions and
verify the expected outcomes.
Chapter 7

Results and Discussion

In this section, we present the results and discussion of the various modules of the
Time Interiors application. The testing outcomes are structured according to the
use case diagram. Each module’s functionality is thoroughly tested, and the steps
required to perform specific actions within the app are elaborated. Screenshots
are provided to verify the successful execution of these actions.

7.1 Results of App.

7.1.1 Login
User will have to login to use app features, for that user must enter the username
and password. If he does not have an account then he have to register first and
then login.Also he have to enter valid info. Here is login screen:

Figure 7.1: Login

29
Chapter 7. Results 30

7.1.2 Register
User will have to Register first to login and use app features, for that user must
enter the username, email and set a password,all credentials should be valid. If he
already have an account then he will simply login. Here is account register screen:
Results for Register:

Figure 7.2: Registraion credentials

7.1.3 Home Page

After successful login, there would be a home screen displaying catalogs of different
products and different room themes, items, reviews and staff profiles.Following is
the home screen:

Figure 7.3: Home Page

Chapter 7. Results 31

7.1.4 Model Screen

Now it is the turn of model screen which is opened from the bottom navigation bar,
with camera icon. When we click it, there will be a screen allowing user to take or
upload wall image with some instructions.This picture will then be classified using
integrated trained CNN model. If the image will be classified as seepaged wall
then there will be relevant products on the screen and if the wall will be classified
as clear then there will be all relevant products on screen. Here is model screen:

Figure 7.4: Model Screen

7.1.5 Product Description

Each product have some descriptions, article number and price with it which will
appear upon click the product. Following is the description of a product:

Figure 7.5: Product Description

Chapter 7. Results 32

7.1.6 Price Calculator

After selecting an item there will be a screen to calculate the price of that item
according to the quantity that a user will require. For that purpose user will have
to enter the dimensions of his area, from that the total area will be calculated and
multiplied by the price per unit.Hence the total price will be appeared, here are
results for price calculator:

Figure 7.6: Price Calculator

7.1.7 Cart
Now user cart the item and also book order. Results for Add to cart:

Figure 7.7: Cart

7.1.8 Review
Users can review for some item as well as for whole service provided to him from
the company. Results for Reviews:
Chapter 7. Results 33

Figure 7.8: Reviews

7.1.9 Staff Profiles

As the user will physically meet the staff hence there are profiles of workers to
know them in advance. Results for Staff Profiles:

Figure 7.9: Staff

7.1.10 Admin Side

After user side results, here are some admin side results. Admin can view users
their orders,notify them, update them with the latest tracking status of their
booked items. Admin can add and remove products their categories, workers and
may also view the monthly and yearly revenue. Results for Admin Side:

Figure 7.10: Admin Side

Chapter 7. Results 34

Figure 7.11: Admin Side

7.2 Results for Model

We trained different types of Convolutional Neural Networks (CNN) models to
classify the given wall image into Seepage class or Clear class. We made a dataset
for it and divide it in 70 percent training and 30 percent testing data.For this
purpose we used Resnet.50, User defined and MobileNetV2 models and compare
the accuracy of all these models and got the MobilNetV2 best in it. Here is the
results of these models.
7.2.1 Resnet.50:
Resnet.50 model was trained on our dataset but it did not give good results.

Figure 7.12: Resnet.50 Model Results

7.2.2 User Defined:

For this model we set some layers and tried to get required results. The results
were quite better than Resnet.50 model.
Chapter 7. Results 35

Figure 7.13: User Defined Model Results

7.2.3 MobileNetV2:
We trained another CNN model named MobileNetV2, on our dataset and it re-
vealed the best results among all models which we used.

Figure 7.14: MobileNetV2 Model Results

Chapter 8

Conclusions and Future Work

8.1 Conclusion
The Time Interiors app is an effective combination of a new technology and the
requirements of interior design, giving the consumers the ability to choose and
purchase the necessary interior design products. During the course of the project,
we built a responsive mobile application using the Flutter framework with Dart
and designed intuitive user flows and interfaces with Figma while making use of
efficient backend services from Firebase. Moreover, we asked TensorFlow to train
the Convolutional Neural Network (CNN) model i.e. Mobile Net V2 which would
allow the app to identify the material for the wall based on images. In addition
to making it easier and faster for users to select the appropriate material, it also
helps ensure that the users choose materials that are best suited for the condition
in their walls, thus improving the satisfaction of the customers and increasing
the profitability of the company through high order volumes through its online
platform.

8.2 Future Work

Further recommendations for the Time Interiors app can be creation of new mod-
ules and enhancing the efficiency of existing ones. Augment Reality features are
one potential improvement that can enable the users to see what the materials look
before they decide to buy them. This would be more engaging for the audience and
would give them a better sense of interacting with the light show. Another area
for improvement is to try to fine-tune the CNN so as to minimize the computa-
tion time and raise the accuracy of the model, perhaps using more technologically
sophisticated methods or try enlarging the set of training datasets.

36
Appendix A

User Manual

This user manual is designed to provide guidance on how to use the Time Interiors
app effectively. The manual covers all the functionalities available to both the
Admin and End User actors. Each section provides step-by-step instructions,
along with relevant screenshots, to ensure a comprehensive understanding of the
app’s features.
A.0.1 A.1 End User Manual
A.0.1.1 A.1.1 User Registration

1. Open the app and click on the ”Register” button located on the main menu.

2. A registration form will appear. Enter your details, including your name,
email address, and password.

3. Click the ”Submit” button. A message will appear saying, ”Your registration
is done, please confirm your email.”

4. Check your email for a confirmation message from Time Interiors.

5. Open the email and click the ”Verify” button. The system will confirm your
registration by showing a message.

A.0.1.2 A.1.2 Login

1. On the login page, enter your username and password.

2. Click the ”Login” button. If the credentials are correct, you will be navigated
to the home screen.

A.0.1.3 A.1.3 Reset Password

1. On the login page, click the ”Forgot Password” link. A form will appear.

37
Appendix A. User Manual 38

2. Enter your email address and press the ”Reset” button. An email will be
sent to your email address with password reset instructions.

3. Open the received email, and you will find a reset password link. This link
is valid for 1 hour only.

4. Click on the link, and a new page will appear. Enter your new password,
confirm it, and click ”Reset”.

5. A success message will appear. You can now log in using the new password.

A.0.1.4 A.1.4 Browse and Select Materials

1. From the home screen, navigate to the catalog of interior design materials.

2. Browse through categories such as wallpaper, wall paneling, and wooden

flooring.

3. Select a material to view detailed information and pricing.

A.0.1.5 A.1.5 Upload Wall Image for Analysis

1. Click on the ”Analyze Wall” button from the main menu.

2. Use your device’s camera to take a picture of the wall or upload an existing
image.

3. The app will analyze the image and suggest suitable materials based on the
wall condition.

A.0.1.6 A.1.6 Place an Order

1. Select the materials you wish to purchase and add them to your cart.

2. Click on the cart icon to review your selections.

3. Proceed to checkout, enter your shipping details, and choose a payment

method.

4. Confirm your order. You will receive an order confirmation message and
email.

A.0.1.7 A.1.7 Submit User Feedback

1. Navigate to the ”Feedback” section from the main menu.

2. Fill out the feedback form with your comments and suggestions.

3. Click the ”Submit” button to send your feedback.

Appendix A. User Manual 39

A.0.2 A.2 Admin Manual

A.0.2.1 A.2.1 Admin Login

1. Access the admin login page and enter your admin credentials.

2. Click the ”Login” button to access the admin dashboard.

A.0.2.2 A.2.2 Manage Users

1. From the admin dashboard, navigate to the ”Manage Users” section.

2. View the list of registered users, with options to edit or delete user accounts.

A.0.2.3 A.2.3 Manage Products

1. Click on the ”Manage Products” tab from the admin dashboard.

2. Add new products by entering details such as name, category, price, and
image.

3. Edit or remove existing products as needed.

A.0.2.4 A.2.4 View Orders

1. Navigate to the ”View Orders” section to see a list of all customer orders.

2. Click on an order to view detailed information, including customer details

and order status.

3. Update the order status as it progresses through fulfillment.

A.0.2.5 A.2.5 View User Feedback

1. Access the ”Feedback” section from the admin dashboard.

2. Review feedback submitted by users, which can include comments, sugges-

tions, and ratings.

A.0.2.6 A.2.6 Add Workers

1. Click on the ”Manage Workers” tab from the admin dashboard.

2. Add new workers by entering their details such as name, contact information,
and role.

3. Edit or remove worker details as needed.

Appendix A. User Manual 40

A.0.2.7 A.2.7 Generate Reports

1. Access the ”Reports” section from the admin dashboard.

2. Generate various reports, such as sales reports, user activity reports, and
product performance reports.

This user manual ensures that both end users and admin users can navigate and
utilize the Time Interiors app effectively, covering all necessary functionalities and
providing clear, step-by-step instructions.
Appendix B

Technical Manual

This manual is intended for programmers and developers who wish to install the
Time Interiors software or make changes to the existing setup.

B.1 Installing Required Software

B.1.1 Installing Visual Studio Code
To install Visual Studio Code, follow these steps:

1. Visit the official Visual Studio Code website at https://code.visualstudio.com/.

2. Click on the ”Download” button for your operating system (Windows, ma-
cOS, or Linux).

3. Once the download is complete, run the installer.

4. Follow the on-screen instructions to complete the installation process.

5. After installation, launch Visual Studio Code.

Figure B.1: Install VSCode

41
Appendix B. Technical Manual 42

B.2 Project Setup

To set up the Time Interiors project, follow these steps:

1. Clone the project repository from GitHub or obtain the project files from
the designated source.

2. Open Visual Studio Code and navigate to the project directory.

3. Install any necessary dependencies or packages using the appropriate com-

mand (e.g., flutter pub get in Terminal).

4. Update the configuration files with the required settings, such as database
connection details and API keys.

5. Run any database migration scripts to initialize the database schema.

B.3 Project Structure

The Time Interiors project follows a structured organization to facilitate easy
development and maintenance. The main components of the project structure
include:

• Controllers: Contains the logic for handling HTTP requests and responses.

• Modules: Organized sets of related functionality, such as user authentica-

tion, product management, and order processing.

• Views: Templates and UI components for rendering the user interface.

• Public: Static assets such as images, stylesheets, and client-side JavaScript

files.

• Config: Configuration files for setting up the environment, database con-

nections, and other application settings.

• Tests: Contains unit tests and integration tests to ensure the reliability of
the codebase.

Programmers can navigate through these directories to locate specific files and
make necessary changes according to the project requirements.
References

[1] Esteban Cumbajin, Nuno Rodrigues, Paulo Costa, Rolando Miragaia, Luı́s
Frazão, Nuno Costa, Antonio Fernández-Caballero, Jorge Carneiro, Leire H
Buruberri, and António Pereira. A systematic review on deep learning with
cnns applied to surface defect detection. Journal of Imaging, 9(10):193, 2023.
[2] Prahar M Bhatt, Rishi K Malhan, Pradeep Rajendran, Brual C Shah, Shan-
tanu Thakar, Yeo Jung Yoon, and Satyandra K Gupta. Image-based surface
defect detection using deep learning: A review. Journal of Computing and
Information Science in Engineering, 21(4):040801, 2021.
[3] Qianqian Zhou, Zuxiang Situ, Shuai Teng, and Gongfa Chen. Convolu-
tional neural networks–based model for automated sewer defects detection
and classification. Journal of Water Resources Planning and Management,
147(7):04021036, 2021.
[4] Young-Jin Cha, Wooram Choi, and Oral Büyüköztürk. Deep learning-based
crack damage detection using convolutional neural networks. Computer-Aided
Civil and Infrastructure Engineering, 32(5):361–378, 2017.
[5] John B Butcher, CR Day, JC Austin, PW Haycock, David Verstraeten, and
Benjamin Schrauwen. Defect detection in reinforced concrete using random
neural architectures. Computer-Aided Civil and Infrastructure Engineering,
29(3):191–207, 2014.
[6] Chongzhi Wu, Li Hong, Lin Wang, Runhong Zhang, Samui Pijush, and Wen-
gang Zhang. Prediction of wall deflection induced by braced excavation in
spatially variable soils via convolutional neural network. Gondwana Research,
123:184–197, 2023.
[7] CG Chua and Anthony TC Goh. Estimating wall deflections in deep exca-
vations using bayesian neural networks. Tunnelling and underground space
technology, 20(4):400–409, 2005.
[8] Ruoxu Ren, Terence Hung, and Kay Chen Tan. A generic deep-learning-based
approach for automated surface inspection. IEEE transactions on cybernetics,
48(3):929–940, 2017.
43
References 44

[9] Eleni N Chatzi, Badri Hiriyur, Haim Waisman, and Andrew W Smyth. Ex-
perimental application and enhancement of the xfem–ga algorithm for the
detection of flaws in structures. Computers & Structures, 89(7-8):556–570,
2011.

[10] Tian Wang, Yang Chen, Meina Qiao, and Hichem Snoussi. A fast and robust
convolutional neural network-based defect detection model in product quality
control. The International Journal of Advanced Manufacturing Technology,
94:3465–3471, 2018.