CAPSTONE Final Review Report

SEAT OCCUPANCY DETECTION

2. Students Name and Roll Number:

3. Team No.: 35
 Introduction
Project Overview

● The project aims to develop a system to detect whether the seats are occupied or unoccupied in an environment.
● The prime focus of our project is to implement this in our college library.
● The aim is to achieve real-time detection by harnessing images captured through a camera and leveraging the capabilities of YOLO
computer vision algorithm.

Need Analysis and Scope

● Space Optimization: Libraries often have limited space, efficient space utilization is crucial to accommodate the needs of patrons.
● Enhanced User experience: Our system provides real-time information on seat availability, enabling users to quickly locate vacant seats.
● Data-Driven Decision Making: System generates valuable data on seating patterns, usage trends and peak hours, enabling library administrators to
make informed decisions regarding space allocation, facility planning, and service improvements.
● Accessibility and Inclusivity : For patrons with mobility issues or special needs, knowing accesible seats can greatly enhance their user experience.

Objectives

● Accuracy and Reliability: Develop a system capable of accurately detecting seat occupancy in real time, ensuring reliable information.
● Efficiency: Implement an automated system that streamlines seat management processes, reducing manual effort.
● User Experience Enhancement: Enhance the library experience for patrons by providing easy access to real-time information for seat
availability.
● Integration: Integrate the seat occupancy detection system with existing library management systems or infrastructure.
 Literature Survey
Application of Deep Learning for Object Detection By Ajeet Ram Pathak and Manjush Pandey

● Traditional object detection methods relied on handcrafted features and complex algorithms, limiting their performance in real-world scenarios.
● Deep learning techniques, on the other hand, automatically learn hierarchical representations from raw data, resulting in more robust and
adaptable object detection systems.
● The study discusses popular deep learning architectures like YOLO, Faster R-CNN, and SSD, highlighting their varying designs and trade-offs between
speed and accuracy.

Object Detection using TensorFlow By Yellamma Pachipala ,M Harika ,B Aakanksha ,M Kavitha

● The study emphasizes TensorFlow's rich ecosystem for object detection, which includes libraries, tools, and pre-trained models.
● It simplifies custom model development using TensorFlow's high-level APIs and pre-built architectures like the TensorFlow Object Detection API.
● Through step-by-step guidance and code examples, the study demystifies the process of building, training, evaluating, and deploying object
detection systems, catering to both novices and experts.

Empty Seat Revelation Using Face Detection and Correlation Matching By Sree Sharmila T and Pavithra L K

● The video captured by the cameras is split into frames and analyzed for human presence. Initially, template-based correlation matching identifies
the initial count of seats in the hall
● The paper proposes face detection using the efficient Viola-Jones algorithm and template matching for initial seat count detection, aided by Fast
Fourier Transform (FFT) for reduced time complexity.
● The face detection process involves breaking the input image into sub-windows and subjecting each to haar feature tests, resulting in a single value
obtained from the subtraction of intensities in black and white areas.
 Literature Survey
Frame filtering and Object Detection By Vivian Siahaan And Rismon Hasiholan sianipar

● This project is a Python-based GUI application that leverages Tkinter alongside OpenCV, Pillow, imageio, and Matplotlib libraries for analyzing RGB
histograms of video frames.
● Built within the `AnalyzeHistogramFrame` class, the application provides an interactive interface with buttons and canvases for user interaction.
● Users can also open multiple independent instances of the application for simultaneous analyses. This setup combines real-time video control with
customizable histogram analysis for an intuitive user experience.

Focal Loss for Dense Object Detection By Lin, T. Y., Goyal, P., Girshick, R., He, K., & Dollar, P.

● This paper addresses a significant challenge in object detection: the extreme class imbalance often encountered in dense detection tasks,
where the number of background examples vastly exceeds that of the foreground objects.
● The paper demonstrates that integrating Focal Loss into existing detection frameworks.
● This work highlights the importance of addressing class imbalance in training models for visual recognition tasks and offers a practical
solution that has influenced subsequent research in the field.

Seat Occupancy Detection Using Machine Learning By Kalyan Venkatesh And Swathi Darla

● With Convolution neural network being a beneficent recognizer and processor of the image, this thesis of Seating Occupancy was made
possible.
● Seat occupancy monitoring involves techniques to sense the presence of a human occupant and the algorithm to relay the information and
display the seat is occupied or unoccupied.
● Also YOLO makes the idea of detecting object according to need easy in terms of speed and accuracy.
 Literature Survey
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks By Ren, S., He, K., Girshick, R., & Sun, J.

● This paper presents a significant advancement in the field of object detection by introducing Faster R CNN, a framework that combines region
proposal networks (RPN) with a Fast R-CNN detector.
● The core innovation lies in the RPN, which generates high-quality region proposals directly from the convolutional feature maps of the input
image, effectively eliminating the need for external proposal generation methods that were previously used.
● This integration enables the framework to achieve near real-time performance while maintaining high accuracy.

A computer vision-based proposal for seat occupancy monitoring By José Miguel Seruca Veloso

● This research paper presents a detailed exploration of seat occupancy monitoring systems, specifically focusing on vision-based solutions.
● The system requirements section specifies both functional and non-functional needs, particularly focusing on occupancy detection accuracy
and user interface design.
● A dashboard is integrated to display real-time occupancy data, offering a user-friendly interface for monitoring and data interpretation

YOLOv4: Optimal Speed and Accuracy of Object Detection By Bochkovskiy, Wang, and Liao

● In this paper they introduces YOLOv4, a state-of-the-art object detection model that balances high accuracy with real-time speed, making it
efficient for practical applications.
● This research addresses the challenge of designing an object detection framework that performs well across a wide range of environments
while optimizing computational efficiency, allowing deployment on systems with limited resources like edge devices.
● This paper’s contributions make YOLOv4 a popular choice for real-time object detection tasks, particularly in applications requiring fast
processing and high accuracy, such as surveillance, traffic monitoring, and autonomous driving.
 Literature Survey

Research Gaps

● Research in computer vision for library seat occupancy detection should prioritize enhancing accuracy in diverse environments
● Achieving real-time performance for timely updates to patrons, ensuring scalability to monitor large seating areas efficiently .
● Implementing privacy-preserving techniques to safeguard patron identities while providing valuable occupancy data.
● There is a challenge in accurately determining occupancy when seats are partially occupied or when personal belongings like backpacks are placed on
empty seats.
● Current systems lack research into the long-term maintenance of hardware (e.g. cameras) and the periodic recalibration or retraining of YOLO to
accommodate changes in seating arrangements or environmental conditions.

Research Findings for Existing Literature

● Accuracy Improvement: Studies have proposed various techniques to enhance the accuracy of computer vision algorithms in detecting seat occupancy,
considering factors like varying lighting conditions, occlusions, and diverse seating arrangements.
● Real-time Performance Optimization: Researchers have explored optimization methods to improve the speed of image processing and analysis without
compromising accuracy, enabling real-time updates on seat availability for library patrons.
● Scalability Solutions: Literature suggests scalable approaches to efficiently monitor a large number of seats simultaneously in libraries, addressing
challenges related to space size and seating density.
● Integration with Existing Systems: Seamless integration with existing library management systems or infrastructure is essential to facilitate
smooth operation and minimize disruption to existing workflows.
 Proposed Approach
Flow Chart : depicting the proposed concept
 Proposed Approach
Class Diagram
 Proposed Approach
Sequence Diagram : detailed Schematic
Prospective Learning of the Project

● Understanding of Computer Vision Principles: We gained a comprehensive understanding of computer vision principles,
including image processing, feature extraction, object detection, and recognition, through hands-on implementation in a
real-world application.

● Machine Learning Techniques: We learned and applied machine learning techniques for pattern recognition,
classification, and regression to develop models for seat occupancy detection based on visual data.

● Data Processing and Analysis Skills: We acquired skills in processing and analyzing data collected by sensors and cameras,
including techniques from digital signal processing and data structures and algorithms.

● System Design and Integration: We learned how to design and integrate components of a seat occupancy detection
system, including hardware sensors, computer vision algorithms, network communication, and database management.

● Wireless Communication and Networking: We gained knowledge of wireless communication principles, protocols, and
technologies, as well as network security concepts, relevant for implementing communication interfaces in the system.

● Database Management and Querying: We learned database design principles and SQL querying techniques for storing,
managing, and retrieving occupancy data collected by the system.
IEEE Standards to be used

● IEEE 802.3 (Ethernet) for wired communication (Distance upto 100 meters)
● IEEE 802.1X for network access control (Range depends on underlying Ethernet or Wi-Fi network).
● IEEE 802.3af/at (PoE) for power delivery over Ethernet (Range of 100 meters per segment)
● IEEE 802.15.4 for LR-WPANs (Range close to 10-30 meters).
● IEEE 1588 (Precision Time Protocol) for clock synchronization (Range upto 100 meters for Ethernet segments).
● IEEE 1901 for Powerline Communication (Range upto 500 meters).

UG Courses referred

● Embedded Systems
● Object Oriented Programming
● Artificial intelligence
● Data Structures And Algorithms
● Database Management Systems
 Outcomes of the Project
. Graphs :
 Outcomes of the Project
. Simulation Results :
 Outcomes of the Project
. Graphical User Interface : Authentication (Login and Registration Page)
 Outcomes of the Project
. Graphical User Interface : Occupancy Output
Project Timeline/Gantt Chart
Thank you