Object Detection and Recognition

for Retail Inventory Management

By

Talat Mehedi Anik - Elias Islam Jewel - MD Jubayer - MD Adnan

Rayhan Shoran - Badhan Rajib

Submitted in partial fulfilment of the requirements

of the degree of Bachelor of Science in Computer Science and Engineering

April 20, 2024

Department of Computer Science and Engineering

United International University
 l

Abstract

In today’s dynamic marketplace, efficient inventory management stands as a critical

factor for businesses to thrive. Traditional methods often fall short in adapting to the com-
plex and evolving demands of modern supply chains. This paper introduces DeepLearning
Inventory, a pioneering framework that harnesses the power of deep learning to revolu-
tionize inventory management. By integrating advanced neural network architectures with
comprehensive data analysis techniques, DeepLearning Inventory offers unparalleled accu-
racy in demand forecasting, optimal inventory level determination, and adaptive replen-
ishment strategies. Through extensive experimentation and real-world case studies, we
demonstrate the efficacy of our approach in enhancing operational efficiency, minimizing
stockouts, and maximizing profit margins. Our findings underscore the transformative po-
tential of deep learning in reshaping traditional inventory management paradigms, paving
the way for a more agile and responsive supply chain ecosystem.

i
 Acknowledgements

This work would have not been possible without the input and support of Talat Mehedi
Anik - Elias Islam Jewel - MD Jubayer - MD Adnan Rayhan Shoran - Badhan Rajib. We
would like to thank my academic advisors, Muzahid sir and Ashiqur sir.

ii
 Publication List

The main contributions of this research are either published or accepted or in prepa-
ration in journals and conferences as mentioned in the following list:

Journal Articles
1.

Conference Papers
1.

Additional Publications
Following is the list of relevant publications published in the course of the research that
is not included in the thesis:

1.

iii
Table of Contents

Table of Contents v

List of Figures vi

List of Tables vii

1 Introduction 1
1.1 Project Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Project Outcome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Organization of the Report . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Background 10
2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Similar Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.2 Related Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Gap Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Project Design 15
3.1 Requirement Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.1 Functional and Nonfunctional Requirements . . . . . . . . . . . . . . 15
3.1.2 Context Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.3 Data Flow Diagram Level 1 . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.4 UI Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Detailed Methodology and Design . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Project Plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4 Task Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

iv
Table of Contents Table of Contents

4 Implementation and Results 16

4.1 Environment Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Testing and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5 Standards and Design Constraints 18

5.1 Compliance with the Standards . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.1.1 Software Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.1.2 Hardware Standards . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.1.3 Communication Standards . . . . . . . . . . . . . . . . . . . . . . . . 18
5.2 Design Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.2.1 Economic Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2.2 Environmental Constraint . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2.3 Ethical Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2.4 Health and Safety Constraint . . . . . . . . . . . . . . . . . . . . . . 19
5.2.5 Social Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2.6 Political Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2.7 Sustainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.3 Cost Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.4 Complex Engineering Problem . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.4.1 Complex Problem Solving . . . . . . . . . . . . . . . . . . . . . . . . 19
5.4.2 Engineering Activities . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

6 Conclusion 21
6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.2 Limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

References 22

v
List of Figures

1.1 Our using process for this project . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 System Administrator and User authentication . . . . . . . . . . . . . . . . 4
1.3 How To Updathe and Add Stock Item . . . . . . . . . . . . . . . . . . . . . 5
1.4 Enhanch Customer Satisfaction . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Data Collection Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.6 Stock Denominations And Issues Required . . . . . . . . . . . . . . . . . . . 7
1.7 ORGANIZATION ,REPORT and feature Extraction . . . . . . . . . . . . . 9

2.1 Determination,Analysis and Discussion . . . . . . . . . . . . . . . . . . . . . 13

2.2 Comparison Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

vi
List of Tables

5.1 Mapping with complex problem solving. . . . . . . . . . . . . . . . . . . . . 19

5.2 Mapping with complex engineering activities. . . . . . . . . . . . . . . . . . 20

vii
Chapter 1

Introduction

In modern inventory management systems, the accurate identification and tracking of

inventory items play a pivotal role in ensuring operational efficiency, minimizing stockouts,
reducing inventory holding costs, and enhancing overall profitability. Traditional methods
of inventory management often rely on manual counting and barcode scanning, which
can be time-consuming, error-prone, and inefficient, especially in large-scale warehouse
environments.
To address these challenges, the integration of object detection and recognition tech-
nologies has emerged as a promising solution. Object detection and recognition leverage
advanced machine learning and computer vision techniques to automatically detect, clas-
sify, and track inventory items based on visual data captured by cameras or sensors. By
harnessing the power of artificial intelligence (AI) and deep learning algorithms, these
technologies enable real-time, automated inventory monitoring, replenishment, and opti-
mization.
This project aims to develop and implement an object detection and recognition sys-
tem tailored specifically for inventory management applications. By accurately identifying
and tracking inventory items in real-time, the system seeks to streamline inventory opera-
tions, improve inventory accuracy, enhance supply chain visibility, and ultimately optimize
inventory management processes for businesses across various industries.
Through the utilization of state-of-the-art object detection models, such as Faster R-
CNN, YOLO (You Only Look Once), or SSD (Single Shot Multibox Detector), combined
with innovative data preprocessing techniques and robust integration with existing in-
ventory management systems, this project endeavors to deliver a scalable, reliable, and
cost-effective solution for modernizing inventory management practices.
With the successful implementation of this object detection and recognition system,
businesses can expect to achieve significant improvements in inventory accuracy, opera-
tional efficiency, and overall competitiveness in today’s dynamic marketplace.

1
1.1. Project Overview Chapter 1. Introduction

Figure 1.1: Our using process for this project

1.1 Project Overview

The paper discusses the emergence of an IoT-based Intelligent Fitness System, aiming to
revolutionize the fitness industry by providing personalized monitoring and guidance to
exercisers. It highlights the growing interest in fitness among urban white-collar workers
and the significant role of technology in driving this trend. By leveraging sensors embedded
in fitness apparatuses and employing machine learning algorithms, the system analyzes
workout data to offer tailored recommendations for users. Implementation details, includ-
ing the use of Django and Nginx for the web server, are presented, showcasing the modern
web technologies utilized in system development. The paper concludes by emphasizing
the potential of combining artificial intelligence and IoT technology to promote healthier
lifestyles through intelligent fitness monitoring and guidance systems.

1.2 Motivation
1. *Operational Efficiency:* Manual inventory management processes are labor-intensive
and prone to errors, leading to inefficiencies in warehouse operations. By implementing
object detection and recognition, businesses can automate inventory tracking and reduce
the need for manual intervention, thereby improving operational efficiency.
2. *Real-time Visibility:* Traditional inventory management methods may not pro-
vide real-time visibility into inventory levels, leading to stockouts or overstock situations.
Object detection and recognition enable real-time monitoring of inventory levels, allowing
businesses to make timely decisions and optimize inventory levels.
3. *Cost Reduction:* Inaccurate inventory management can result in increased holding
costs, stockouts, and lost sales opportunities. By accurately tracking inventory items, busi-

2
1.3. Objectives Chapter 1. Introduction

nesses can reduce holding costs, minimize stockouts, and avoid excess inventory, leading
to cost savings.
4. *Enhanced Customer Satisfaction:* Timely and accurate inventory management
ensures that products are available when customers need them, leading to improved cus-
tomer satisfaction and loyalty. Object detection and recognition help businesses meet
customer demands more efficiently and effectively.
5. *Competitive Advantage:* In today’s competitive business environment, companies
are constantly seeking ways to gain a competitive edge. Implementing advanced tech-
nologies like object detection and recognition for inventory management can differentiate
businesses from competitors and position them as industry leaders.
6. *Data-driven Decision Making:* Object detection and recognition provide valuable
data insights into inventory trends, demand patterns, and supply chain dynamics. By
leveraging this data, businesses can make informed decisions, optimize inventory levels,
and identify opportunities for process improvement and innovation.
7. *Adaptation to Market Trends:* With the rise of e-commerce, omnichannel re-
tailing, and dynamic consumer preferences, businesses need agile and adaptive inventory
management solutions. Object detection and recognition enable businesses to adapt to
changing market trends and customer demands more effectively.
8. *Compliance and Risk Management:* For industries with strict regulatory require-
ments, such as pharmaceuticals or food and beverage, accurate inventory management is
essential for compliance and risk management. Object detection and recognition help busi-
nesses maintain compliance with regulations and mitigate risks associated with inventory
discrepancies or shortages.
In summary, the motivation for implementing object detection and recognition for in-
ventory management lies in the desire to improve operational efficiency, enhance real-time
visibility, reduce costs, enhance customer satisfaction, gain a competitive advantage, make
data-driven decisions, adapt to market trends, and ensure compliance and risk manage-
ment. By addressing these motivations, businesses can unlock significant benefits and
drive success in today’s dynamic marketplace.

1.3 Objectives
1. *Develop an Accurate Object Detection System:* The primary objective of this project
is to develop an accurate and robust object detection system capable of identifying and
recognizing inventory items in various environmental conditions and settings.
2. *Enable Real-time Inventory Tracking:* Implement real-time inventory tracking ca-
pabilities to monitor the movement and location of inventory items continuously, providing
up-to-date information on inventory levels and statuses.
3. *Automate Inventory Management Processes:* Automate inventory management
processes, such as inventory counting, tracking, and replenishment, to reduce manual
labor, minimize errors, and improve operational efficiency.

3
1.3. Objectives Chapter 1. Introduction

Figure 1.2: System Administrator and User authentication

4. *Enhance Inventory Visibility:* Enhance inventory visibility by providing stake-

holders with comprehensive insights into inventory levels, trends, and patterns, enabling
informed decision-making and strategic planning.
5. *Optimize Inventory Levels:* Optimize inventory levels by accurately predicting
demand, identifying slow-moving or obsolete items, and minimizing stockouts or overstock
situations to reduce holding costs and maximize profitability.
6. *Improve Supply Chain Efficiency:* Improve supply chain efficiency by streamlining
inventory operations, reducing lead times, and enhancing collaboration between suppliers,
manufacturers, distributors, and retailers.
7.*Enhance Customer Satisfaction:* Enhance customer satisfaction by ensuring prod-
uct availability, minimizing order fulfillment times, and providing accurate and timely
information on product availability and delivery status.
8. *Enable Seamless Integration:* Enable seamless integration with existing inventory
management systems, ERP (Enterprise Resource Planning) systems, and other business
applications to facilitate data exchange and workflow automation.
9. *Ensure Scalability and Adaptability:* Design the object detection system to be
scalable and adaptable to accommodate changing business requirements, fluctuations in
inventory volumes, and technological advancements over time.
10. *Comply with Regulatory Requirements:* Ensure compliance with regulatory
requirements and industry standards related to inventory management, data privacy, and
security to mitigate risks and maintain trust with stakeholders.
11. *Provide Training and Support:* Provide training and support to users and stake-
holders to ensure the successful adoption and utilization of the object detection system,
including user training, documentation, and ongoing technical support.
By achieving these objectives, the project aims to revolutionize inventory management
practices, improve operational efficiency, enhance inventory visibility, optimize inventory

4
1.4. Methodology Chapter 1. Introduction

Figure 1.3: How To Updathe and Add Stock Item

Figure 1.4: Enhanch Customer Satisfaction

levels, and ultimately drive business success in today’s competitive marketplace.

1.4 Methodology
Data Collection and Annotation: Gather a diverse dataset of images or videos representing
inventory items in different settings. Annotate the dataset with bounding boxes or masks
to indicate the location and type of inventory items. Preprocessing and Augmentation:
Clean and preprocess the collected data, including resizing, cropping, and normalization.
Augment the dataset to increase diversity and robustness, using techniques like rotation,
flipping, and color augmentation. Model Selection and Training: Choose a suitable object
detection model architecture, such as Faster R-CNN or YOLO. Train the selected model
on the annotated dataset using a deep learning framework like TensorFlow or PyTorch.
Model Evaluation and Optimization: Evaluate the trained model’s performance using
metrics like precision, recall, and mean Average Precision (mAP). Fine-tune the model
and hyperparameters to improve performance on specific inventory items and environmen-
tal conditions. Deployment Architecture: Design the deployment architecture, including
hardware components (e.g., cameras, sensors) and software components (e.g., edge de-

5
1.5. Project Outcome Chapter 1. Introduction

Figure 1.5: Data Collection Model

vices, cloud servers). Real-time Object Detection Pipeline: Implement a real-time object
detection pipeline to capture live video streams from cameras or sensors. Process the video
streams through the trained model to detect and recognize inventory items in real-time.
Integration with Inventory Management System: Develop APIs or interfaces to integrate
the object detection system with the existing inventory management system. Enable
seamless communication between the object detection system and inventory databases for
inventory tracking and management. Testing and Validation: Conduct extensive testing
and validation of the deployed system in simulated and real-world environments. Verify
the accuracy, reliability, and scalability of the object detection and recognition system.
Performance Monitoring and Maintenance: Implement monitoring tools to track system
performance metrics, such as detection accuracy and latency. Perform regular maintenance
and updates to ensure the system remains effective and up-to-date. Diagrams:
i. Data Collection and Annotation Process: ii. Model Training Pipeline: iii. Real-time
Object Detection Pipeline: iv. Integration with Inventory Management System: These
diagrams visually represent the steps involved in each stage of the methodology, providing
a clear understanding of the process for implementing object detection and recognition for
inventory management.

1.5 Project Outcome

The proposed outcome of this project is the development and implementation of an ad-
vanced object detection and recognition system for inventory management. The system
will address the key challenges faced by traditional inventory management processes and
deliver the following outcomes:
1. *Accurate and Reliable Inventory Tracking:* The system will accurately identify
and recognize inventory items across various environmental conditions, ensuring reliability

6
1.5. Project Outcome Chapter 1. Introduction

Figure 1.6: Stock Denominations And Issues Required

and minimizing errors in inventory tracking.

2. *Real-time Inventory Monitoring:* Implementing real-time inventory tracking capa-
bilities will enable continuous monitoring of inventory movement and locations, providing
up-to-date information on inventory levels and statuses.
3. *Automated Inventory Management Processes:* Streamlining inventory manage-
ment tasks such as counting, tracking, and replenishment will reduce manual labor, mini-
mize errors, and enhance operational efficiency.
4. *Enhanced Inventory Visibility:* Providing stakeholders with comprehensive in-
sights into inventory levels, trends, and patterns will enable informed decision-making
and strategic planning.
5. *Seamless Integration with Existing Systems:* Facilitating seamless integration with
existing inventory systems, ERP platforms, and other business applications will streamline
data exchange and workflow automation.
6. *Scalable and Adaptable System Design:* Creating a system that can scale and
adapt to evolving business requirements, fluctuations in inventory volumes, and techno-
logical advancements will ensure long-term viability and sustainability.
7. *Compliance and Security Assurance:* Ensuring adherence to regulatory require-
ments and industry standards related to inventory management, data privacy, and security
will mitigate risks and maintain stakeholders’ trust.
*Validation of Results:*
1. *Accuracy Testing:* Evaluate the accuracy of the object detection and recognition
system by comparing the detected inventory items against manually verified ground truth
data.
2. *Real-time Monitoring:* Validate the real-time inventory monitoring capabilities

7
1.6. Organization of the Report Chapter 1. Introduction

by conducting live tests in a simulated or real-world environment, monitoring inventory

movement and locations in real-time.
3. *Process Automation:* Measure the reduction in manual labor and errors by com-
paring the efficiency and accuracy of automated inventory management processes against
manual methods.
4. *Inventory Visibility Analysis:* Assess the effectiveness of the system in providing
comprehensive insights into inventory levels, trends, and patterns by analyzing reports
and dashboards generated by the system.
5. *Integration Testing:* Validate the seamless integration with existing systems by
conducting integration tests to ensure smooth data exchange and workflow automation.
6. *Scalability Testing:* Evaluate the scalability and adaptability of the system by
testing its performance under varying loads, inventory volumes, and business scenarios.
7. *Compliance and Security Audits:* Conduct compliance audits and security assess-
ments to ensure that the system meets regulatory requirements and industry standards
related to inventory management, data privacy, and security.
By rigorously validating the results through these methods, we can ensure that the
proposed outcome meets the objectives of the project and delivers tangible benefits to the
organization.

1.6 Organization of the Report

The project will be organized into several distinct phases to ensure thorough exploration
and analysis of the application of machine learning techniques in vehicle fitness evaluation:

Literature Review: 1.Inventory Management using Machine Learning

Inventory management is crucial for small/medium-sized businesses, requiring signif-
icant investment in both money and skilled labor. E-commerce giants utilize Machine
Learning models for inventory maintenance based on demand. Extending inventory man-
agement as a service to smaller businesses can enhance sales and predict product demand.
Demand forecasting determines how much stock a company should keep to meet demands,
improving customer experience and reducing costs. Challenges in building an inventory
system include designing predictive models and making strategic decisions about inventory
levels.[1]
2,Automation Of Inventory Management Process
The Ugandan government collects revenue through various taxes, including trading
license fees, delegated to local governments. Victoria Insglass Limited, awarded the con-
tract for trading license fee collection in Kampala, manages its licenses manually, leading
to inventory management challenges. The project aims to automate inventory manage-
ment for Victoria Insglass Ltd, ensuring timely decision-making and customer satisfaction.
Literature review underscores the importance of Management Information Systems (MIS)
and automated inventory control systems for efficient business management.[2]

8
1.6. Organization of the Report Chapter 1. Introduction

Figure 1.7: ORGANIZATION ,REPORT and feature Extraction

3.Applications of Artificial Intelligence in Inventory Management: A Systematic Re-

view of the Literature
Today’s customers seek products globally, desiring them promptly, affordably, and of
high quality. Efficient supply chains, crucial for meeting these demands, rely on effective
inventory management. Singh and Verma define inventory management as a process min-
imizing inventory investment while balancing supply and demand. It involves planning
and controlling materials throughout the supply chain, addressing issues like estimating re-
quirements, ordering frequency, and safety stock levels. Uncertainties in demand forecasts
during lead time are crucial to manage, as they guide production planning and opera-
tional decisions. Accurate demand forecasts are vital, as they impact service levels and
costs. Artificial intelligence, particularly machine learning and deep learning, enhances
inventory management by analyzing large datasets and improving forecasting accuracy.
Integration of AI with inventory management reduces operational costs and enhances
flexibility, providing faster response times and contextual information to meet customer
needs efficiently.[3]
4.COMPUTER VISION (AI) BASED RETAILER SHELVES MONITORING SYS-
TEM TO NOTIFY EMPTY SHELVES — HARISH KUMAR
5, Optimizing Supply Chain Processes through Deep learning Algorithms: A Manage-
rial Approach. 6.HYBRID APPROACH FOR SHELF MONITORING AND PLANOGRAM
COMPLIANCE (HYB-SMPC) IN RETAILS USING DEEP LEARNING AND COM-
PUTER VISION 15 JUN 2022 — SADDAF RUBAB link:: https://www.hindawi.com/journals/mpe/2022/49

9
Chapter 2

Background

In traditional inventory management systems, manual counting and barcode scanning

methods are commonly used for tracking and managing inventory items. However, these
methods are often labor-intensive, time-consuming, and prone to errors, especially in large-
scale warehouse environments. Additionally, they may not provide real-time visibility into
inventory levels, leading to inefficiencies, stockouts, and increased holding costs.
To address these challenges, businesses are increasingly turning to advanced technolo-
gies such as object detection and recognition. Object detection and recognition leverage
the power of machine learning and computer vision to automatically identify, classify, and
track inventory items based on visual data captured by cameras or sensors. By harnessing
the capabilities of artificial intelligence (AI) and deep learning algorithms, these technolo-
gies enable real-time, automated inventory monitoring, replenishment, and optimization.
The need for accurate and efficient inventory management has become even more
critical in today’s fast-paced and competitive business landscape. Businesses are facing
growing demands for improved inventory accuracy, operational efficiency, and supply chain
visibility. Furthermore, the rise of e-commerce and omnichannel retailing has increased
the complexity of inventory management, requiring agile and adaptive solutions to meet
customer expectations and market demands.
Against this backdrop, the implementation of object detection and recognition for in-
ventory management offers significant potential benefits. By accurately identifying and
tracking inventory items in real-time, businesses can streamline inventory operations, re-
duce stockouts, minimize holding costs, and enhance overall profitability. Moreover, these
technologies enable businesses to gain valuable insights into inventory trends, demand
patterns, and supply chain dynamics, empowering informed decision-making and strategic
planning.
In light of these considerations, the development and deployment of an object detection
and recognition system tailored specifically for inventory management purposes present a
compelling opportunity for businesses to modernize their inventory management practices,
improve operational efficiency, and gain a competitive edge in the marketplace.

10
2.1. Preliminaries Chapter 2. Background

2.1 Preliminaries
In this section, you have to provide the necessary background knowledge to understand
the rest of the report [1].

2.2 Literature Review

A novel solution to the challenges faced in retail management, particularly in shelf mon-
itoring and planogram compliance. By leveraging deep learning and computer vision
techniques, the proposed Hyb-SMPC approach aims to automate the process of detecting
fine-grained stock-keeping units (SKUs) on shelves and verifying planogram compliance.
Highlighting the importance of product availability and arrangement for customer satisfac-
tion and sales, the paper emphasizes the limitations of manual approaches and the need
for automated solutions. The proposed approach compares three deep learning models
(YOLO V4, YOLO V5, and YOLOR) for SKU detection, aiming to achieve accurate re-
sults. Through a comprehensive review of related work and contributions of the research,
the paper underscores the significance of the proposed approach in optimizing retail opera-
tions and improving customer experience.[1] Deep learning can automate shelf monitoring
and planogram compliance in retail stores. By using object detectors to identify and
count products on shelves, and comparing the layout with a predefined planogram (either
through color matching or JSON data analysis), the system can generate real-time reports
on product availability and planogram adherence, improving inventory management and
potentially boosting customer satisfaction.[2] The paper provides a comprehensive review
of the use of deep learning in inventory forecasting, highlighting its potential benefits
for businesses in terms of data-driven decision-making and improved supply chain opera-
tions. It emphasizes the importance of systematically evaluating existing approaches and
identifying areas for future research. The literature review discusses the significance of
inventory forecasting in business operations and identifies key challenges such as demand
variability and supply chain disruptions.[3]Overall, the paper underscores the importance
of deep learning in enhancing inventory forecasting accuracy and efficiency, while also
acknowledging challenges such as the need for large volumes of data and computational
resources. It calls for further research to address these challenges and advance the field, ul-
timately contributing to improved business management and accounting practices.[4]The
paper delves into the application of deep learning algorithms in optimizing supply chain
processes from a managerial perspective, addressing the complexities of modern supply
chains and the limitations of traditional optimization techniques in adapting to uncer-
tainties. Deep learning, as a subset of artificial intelligence, is highlighted for its ability
to uncover patterns from vast datasets, aiding in areas like demand forecasting, inven-
tory management, route optimization, and risk mitigation. Through techniques such as
neural networks and convolutional neural networks, managers can enhance operational
efficiency, streamline logistics, and identify disruptions proactively.[5]The Ugandan gov-

11
2.2. Literature Review Chapter 2. Background

ernment collects revenue through various taxes, including trading license fees, delegated
to local governments. Victoria Insglass Limited, awarded the contract for trading license
fee collection in Kampala, manages its licenses manually, leading to inventory manage-
ment challenges. The project aims to automate inventory management for Victoria Ins-
glass Ltd, ensuring timely decision-making and customer satisfaction. Literature review
underscores the importance of Management Information Systems (MIS) and automated
inventory control systems for efficient business management [6] However, challenges such
as data quality issues and organizational readiness are acknowledged, emphasizing the
importance of collaboration between data scientists, IT professionals, and supply chain
managers for successful implementation. Overall, the paper underscores the transforma-
tive potential of deep learning in supply chain optimization and provides practical insights
for managers aiming to leverage this technology for organizational success. Problem State-
ment: In traditional inventory management systems, businesses often face challenges in
accurately forecasting demand, optimizing inventory levels, and efficiently replenishing
stock. These challenges can lead to overstocking, stockouts, increased holding costs, and
lost sales opportunities. Moreover, manual or rule-based methods for inventory manage-
ment lack the adaptability and scalability needed to handle dynamic and complex supply
chain environments. Objective: The objective of this research is to develop an inventory
management system leveraging deep learning techniques to address the shortcomings of
traditional methods. By harnessing the power of deep learning algorithms, we aim to en-
hance demand forecasting accuracy, optimize inventory replenishment decisions, minimize
stockouts and overstocking, and ultimately improve overall supply chain efficiency. Key
Challenges:
1. Demand Forecasting Accuracy: Conventional forecasting methods often struggle
to capture the complex patterns and dependencies present in historical sales data. Deep
learning models need to overcome these challenges by effectively capturing nonlinear re-
lationships and temporal dependencies in demand data. 2. Dynamic Inventory Opti-
mization: Traditional inventory optimization techniques rely on static rules or heuristics,
which may fail to adapt to changing demand patterns and supply chain dynamics. Deep
learning-based approaches must dynamically adjust inventory levels based on real-time
demand signals, supply chain disruptions, and other contextual factors. 3. Data Quality
and Availability: Deep learning models require large volumes of high-quality historical
data for training. However, inventory data may suffer from noise, missing values, and
inconsistencies, posing challenges for model training and deployment. 4. Interpretability
and Explainability: Deep learning models are often perceived as black-box algorithms,
making it challenging to interpret their decisions and recommendations. Ensuring trans-
parency and explainability in the inventory management system is crucial for gaining
stakeholders’ trust and acceptance. Proposed Approach: We propose to develop a deep
learning-based inventory management system comprising several components: • Demand
Forecasting Module: Utilizing recurrent neural networks (RNNs), convolutional neural
networks (CNNs), or transformer architectures to capture temporal dependencies and

12
2.2. Literature Review Chapter 2. Background

Figure 2.1: Determination,Analysis and Discussion

nonlinear patterns in demand data. • Inventory Optimization Module: Employing rein-

forcement learning (RL), deep Q-networks (DQN), or other deep reinforcement learning
techniques to dynamically optimize inventory replenishment decisions based on real-time
demand forecasts and inventory holding costs. • Data Preprocessing and Feature Engineer-
ing: Cleaning, preprocessing, and feature engineering techniques to handle missing data,
outliers, and extract relevant features from historical inventory and sales data. • Model
Interpretability and Visualization: Integrating techniques such as attention mechanisms,
saliency maps, and model-agnostic interpretability methods to enhance the transparency
and interpretability of the deep learning models. Expected Outcomes: By implementing
the proposed inventory management system, we expect to achieve the following outcomes:
• Improved demand forecasting accuracy, leading to reduced forecasting errors and better
alignment of inventory levels with actual demand. • Optimized inventory replenishment
decisions, resulting in reduced holding costs, minimized stockouts, and improved service
levels. • Enhanced agility and responsiveness to dynamic market conditions and supply
chain disruptions. • Increased operational efficiency and cost savings through automation
and optimization of inventory management processes.

2.2.1 Similar Applications

Put a summary of similar web applications, mobile apps similar to your work.

13
2.3. Gap Analysis Chapter 2. Background

Figure 2.2: Comparison Table

2.2.2 Related Research

Here report the summary of the investigation of the research literature.

2.3 Gap Analysis

Here summarise the gap where you intend to work.

2.4 Summary

14
Chapter 3

Project Design

[Must be present in FYDP-1 Report and also in Final Report]

Every chapter should start with 1-2 sentences on the outline of the chapter.

3.1 Requirement Analysis

3.1.1 Functional and Nonfunctional Requirements

3.1.2 Context Diagram

3.1.3 Data Flow Diagram Level 1

3.1.4 UI Design

3.2 Detailed Methodology and Design

You have to mention alternate solutions that you have considered. Why you have selected
the specific solution, etc.

3.3 Project Plan

3.4 Task Allocation

3.5 Summary

15
Chapter 4

Implementation and Results

graphicx [export]adjustbox
[Must be present in Final Report. Incomplete version might be included in FYDP-1
Report, however it is optional.]
Every chapter should start with 1-2 sentences on the outline of the chapter.

16
4.1. Environment Setup Chapter 4. Implementation and Results

4.1 Environment Setup

4.2 Testing and Evaluation

4.3 Results and Discussion

4.4 Summary

17
Chapter 5

Standards and Design Constraints

[Must be present in FYDP-1 Report and also in Final Report]

Every chapter should start with 1-2 sentences on the outline of the chapter.

5.1 Compliance with the Standards

Only mention the standards that are related to your project. This list is not complete. For
each of the standards discuss the alternates with pros and cons and rationale of selection.

5.1.1 Software Standards

5.1.2 Hardware Standards

5.1.3 Communication Standards

5.2 Design Constraints

Only mention the constraints that are related to the design of your project. This list is
not complete.

18
5.3. Cost Analysis Chapter 5. Standards and Design Constraints

5.2.1 Economic Constraint

5.2.2 Environmental Constraint

5.2.3 Ethical Constraint

5.2.4 Health and Safety Constraint

5.2.5 Social Constraint

5.2.6 Political Constraint

5.2.7 Sustainability

5.3 Cost Analysis

Provide a cost analysis in terms of budget required and revenue model. In case of budget,
you must show an alternate budget and rationales.

5.4 Complex Engineering Problem

5.4.1 Complex Problem Solving
In this section, provide a mapping with problem solving categories. For each mapping add
subsections to put rationale (Use Table 5.1). For P1, you need to put another mapping
with Knowledge profile and rational thereof.

Table 5.1: Mapping with complex problem solving.

P1 P2 P3 P4 P5 P6 P7
Dept of Range Depth of Familiarity Extent of Extent Inter-
Knowl- of Con- Analysis of Issues Applicable of Stake- dependence
edge flicting Codes holder
Require- Involve-
ments ment
√ √

5.4.2 Engineering Activities

In this section, provide a mapping with engineering activities. For each mapping add
subsections to put rationale (Use Table 5.2).

5.5 Summary

19
5.5. Summary Chapter 5. Standards and Design Constraints

Table 5.2: Mapping with complex engineering activities.

A1 A2 A3 A4 A5
Range of re- Level of Interac- Innovation Consequences Familiarity
sources tion for society and
environment
√ √

20
Chapter 6

Conclusion

[Must be present in FYDP-1 Report and also in Final Report. Might be incomplete in
FYDP-1 Report.]
Every chapter should start with 1-2 sentences on the outline of the chapter.

6.1 Summary

6.2 Limitation

6.3 Future Work

21
References

[1] Jon Kleinberg and Eva Tardos. Algorithm design. Pearson Education India, 2006.

22