ENHANCING ROAD SAFETY WITH THE LATEST

BREAKTHROUGH: REAL-TIME VEHICLE

CLASSIFICATION, COUNTING AND SPEED ESTIMATION
USING YOLO V8N AND DEEP SORT ALGORITHM

PROJECT REPORT

Submitted by

JEEVITHA.K Register No.: 231723

Under the guidance of

Mrs.DR.S.S. BOOMIGA

in partial fulfillment of the requirements for the degree

of

MASTER OF TECHNOLOGY

DEPARTMENT OF ARTIFICIAL INTELLIGENCE AND DATA SCIENCE

SRI MANAKULA VINAYAGAR ENGINEERING COLLEGE

(An Autonomous Institution)

MADAGADIPET, PUDUCHERRY - 605107

MARCH 2024

i
 DEPARTMENT OF ARTIFICIAL INTELLIGENCE AND DATA SCIENCE

BONAFIDE CERTIFICATE

This is to certify that the project work entitled “ENHANCING ROADSAFETY WITH

THE LATEST BREAKTHROUGH: REAL-TIME VEHICLE CLASSIFICATION,

COUNTING AND SPEED ESTIMATION USING YOLO V8N AND DEEP SORT

ALGORITHM” is a bonafide work done by JEEVITHA.K [231723] in partial fulfillment

of the requirement, for the award of M. Tech Degree in artificial intelligence and data science

by Pondicherry University during the academic year 2023 - 25.

PROJECT GUIDE HEAD OF THE DEPARTMENT

Submitted for the End Semester Practical Examination held on _____________

INTERNAL EXAMINER EXTERNAL EXAMINER

ii
 ACKNOWLEDGEMENT
We are very thankful and grateful to our beloved guide, DR. S.S. BOOMIGA whose
great support in valuable advices, suggestions and tremendous help enabled us in completing
our project. She has been a great source of inspiration to us.

We also sincerely thank our Head of the Department, Dr. J. MADHUSUDANAN

whose continuous encouragement and sufficient comments enabled us to complete our
project report.

We thank all our Staff members who have been by our side always and helped us
with our project. We also sincerely thank all the lab technicians for their help as in the course
of our project development.

We would also like to extend our sincere gratitude and grateful thanks to our Director
cum Principal Dr. V. S. K. VENKATACHALAPATHY for having extended the Research
and Development facilities of the department.

We are grateful to our Founder Chairman Shri. N. KESAVAN. He has been a

constant source of inspiration right from the beginning.

We would like to express our faithful and grateful thanks to our Chairman and
Managing Director Shri. M. DHANASEKARAN for his support.

We would also like to thank our Vice Chairman Shri. S. V. SUGUMARAN, for
providing us with pleasant learning environment.

We would like to thank our Secretary Dr. K. NARAYANASAMY for his support.

We wish to thank our family members and friends for their constant encouragement,
constructive criticisms and suggestions that has helped us in timely completion of this
project.

iii
 Last but not the least; we would like to thank the ALMIGHTY for His grace and
blessings over us throughout the project.

ABSTRACT

The importance of real-time vehicle classification and counting and speed estimation system
based on YOLOv8n is an important tool for monitoring traffic flow on highways. However,
because they are distinct objects from the surroundings, it is still difficult to detect them,
which has an impact on how accurate vehicle counts are. To address this issue, this paper
proposes vision-based vehicle detection, counting and speed estimation system using You
Only Look Once (YOLOv8n) based Deep SORT model for real time vehicle monitoring and
identification of vehicles from video clips. Deep learning based Simple Real time Tracker
(Deep SORT) algorithm is added, which will track actual presence of vehicles from video
frame predicted by YOLOv8n with high accuracy. Two processes are involved in vehicle
counting: first, the recorded video is fed into a deep learning framework based on You Only
Look Once (YOLO) to identify, count, and classify each vehicle. Multi vehicular tracking is
adopted using Deep Sort enhances the tracking performance by associating detected vehicles
across frames, ensuring reliable tracking in crowded and dynamic environments.

Keywords: Vehicle Counting, Object Detection, Tracking, YOLOv8n, Deep SORT

iv
 LIST OF FIGURES
FIGURE NAME OF THE FIGURE PAGE
NO. NO.

1.2 WORKING OF MACHINE LEARNING 3

1.3 TYPES OF MACHINE LEARNING 4

1.3.2 EXAMPLE FOR SUPERVISED LEARNING 5

1.3.7 EXAMPLE FOR UNSUPERVISED LEARNING 10

1.3.8 ILLUSTRATION OF CLUSTERING 12

1.3.9a TYPES OF CLUSTERING 12

1.3.9b PARTITIONING CLUSTERING 13

1.3.9c DENSITY-BASED CLUSTERING 14

1.3.9d DISTRIBUTION MODEL-BASED CLUSTERING 15

1.3.9e HIERARCHIAL CLUSTERING 15

1.4.2 STEPS IN FUZZY C-MEANS CLUSTERING 19

1.5 WORKING OF PYTHON 21

v
 LIST OF ABBREVIATIONS

ML - Machine Learning

FCM - Fuzzy C-Means

DFD - Data Flow Diagram

KNN - K-Nearest Neighbor

EDA - Exploratory Data Analysis

DNN - Deep Neural Network

CNN - Convolutional Neural Network

MLP - Multi-layer Perception

MF - Matrix factorization

CF - Collaborative Filtering

SVM - Support Vector Machine

RMSE - Root Mean Square Error

vi
 TABLE OF CONTENTS

CHAPTER NO. TITLE PAGE NO.

ABSTRACT iv

LIST OF FIGURES v

LIST OF ABBREVIATIONS vi

1. INTRODUCTION
1

1.1 Machine Learning 2

1.2 Working of Machine Learning 3

1.3 Types of Machine learning 4

1.3.1 Supervised Learning 5

1.3.2 Working Of Supervised Learning 5

1.3.3 Advantages Of Supervised Learning 7

1.3.4 Drawbacks Of Supervised Learning 8

1.3.5 Uses Of Supervised Learning 8

1.3.6 Unsupervised Learning 9

1.3.7 Working Of Unsupervised Learning 10

1.3.8 Clustering Algorithm 10

1.3.9 Types Of clustering 12

1.3.10 Benefits Of Clustering 16

1.3.11 Drawbacks Of Clustering 16

1.3.12 Uses Of Clustering 16

1.4 Fuzzy C-Means Clustering 17

1.4.1 Installation and usage 18

vii
 1.4.2 Steps in Fuzzy C-Means 18

1.4.3 Advantages Of Fuzzy C means 19

1.4.4 Disadvantages Of Fuzzy C means 20

1.5 TECHNOLOGY USED 20

1.5.1 Introduction to python

20

1.5.2 Jupyter Notebook 21

1.5.3 Streamlit 21

2. LITERATURE SURVEY
22

2.1 Movie Recommender System Using 22

Collaborative Filtering

2.2 Movie Recommender System Using NLP Tools 23

2.3 Movie Recommendation System Using Weighted 24

Approach
2.4 Movie Recommendation Using YouTube Trailer data 25

2.5 Movie Recommender System Based on Percentage 26

of View
2.6 Movie Recommender System Using critic consensus 27

2.7 Movie Recommender System Using K-Means Clustering 28

2.8 Plot-Topic based Movie Recommendation System using 29

WordNet
2.9 Movie Recommendation Through Sentiment Analysis 30

3. SYSTEM STUDY
31

3.1 EXISTING SYSTEM 31

3.1.1 Issues in the system 31

3.1.2 Solution 31

viii
 3.2 PROPOSED SYSTEM 32

3.2.1 Advantages of Proposed system 32

4. IMPLEMENTATION
33

4.1 MODULE 33

4.2 MODULE DESCRIPTION 33

4.2.1 Data Collection 33

4.2.2 Pre-Processing 33

4.2.3 Clustering 33

4.2.4 Recommender 34

5. SYSTEM REQUIREMENTS
35

5.1 HARDWARE REQUIREMENTS 35

5.2 SOFTWARE REQUIREMENTS 35

6. CONCLUSION AND FUTURE ENHANCEMENT

36

REFERENCES 37

ix
 CHAPTER 1

INTRODUCTION

Real-time vehicle classification, counting, and speed estimation systems are crucial tools that
help transportation authorities and law enforcement agencies monitor traffic conditions and
optimize traffic flow [1][2][3]. Government authorities and private establishment might want
to understand the traffic flowing through a place to better develop its infrastructure for the
ease and convenience of everyone. A road widening project, timing the traffic signals and
construction of parking spaces are a few examples where analysing the traffic is integral to
the project. Traditionally, identification and tracking has been carried out manually. A person
will stand at a point and note the count of the vehicles and their types. Recently, sensors have
been put into use, but they only solve the counting problem. Sensors will not be able to detect
the type of vehicle [10].

In today's rapidly evolving world, effective traffic management and surveillance have become
paramount. Real-time vehicle classification, counting, and speed estimation systems are
crucial tools that help transportation authorities and law enforcement agencies monitor traffic
conditions and optimize traffic flow [1][2][3]. Despite their limitations, video surveillance
systems play a crucial role in monitoring street situations. However, their effectiveness is
largely centered around capturing stable recordings through fixed cameras under simple
conditions. These methods rely on background deduction to identify the stationary elements
of the scene, allowing for mass analysis and tracking of moving vehicles by detecting
bounding boxes. However, these techniques often fail to address various obstacles such as
shadows, unstable cameras, inaccurate camera placement, and complex backgrounds. Smart
cities will greatly benefit from the implementation of this concept [4]. The counting of
vehicles in parking lots can also be implemented using the vehicle counting system. In the
context of traffic management systems, the utilization of the YOLO algorithm for vehicle
counting is of great significance, given its many applications [5] [6]. The current traffic
management system has identified several issues that could be addressed through an analysis
utilizing the YOLO algorithm and deep sorting for vehicle counting [7].

The primary objective of the proposed work is to identify vehicles or objects and monitor
their movements within a specific area of interest. The ultimate aim is to accurately tally the
number of vehicles present on the road. To achieve this, an algorithm called You Only Look
1
Once (YOLO) is put forward as a means of detecting and counting all the objects observed in
the video frames [7][8]. The output of the YOLO algorithm is subsequently inputted into the
Deep SORT tracker, enabling the tracking of objects found within the video frames.

A) YOLO v8: Revolutionizing Object Detection: One of the most important and often used
tasks in computer vision is object detection. YOLO (You Only Look Once) is an
exceptionally powerful algorithm for real-time object detection. YOLOv8n, the latest version,
has taken this technology to new heights. It offers superior accuracy and performance by
dividing images into a grid and predicting bounding boxes and class probabilities within each
grid cell simultaneously.

B) Efficient and Accurate Classification: One significant advantage of YOLOv8n is its

ability to classify vehicles accurately. It can identify and categorize various types of vehicles,
such as cars, trucks, motorcycles, bicycles, and even pedestrians, with remarkable precision.
This reliable classification provides vital insights into traffic density, vehicle composition,
and other crucial parameters for traffic management.

C) Real-Time Processing: The real-time processing capability of YOLOv8n surpasses

traditional methods. With its speed and efficiency, it allows for instant monitoring of multiple
vehicles in a scene simultaneously. This enhanced processing speed is particularly beneficial
for situations.

D) Deep Sort Algorithm Tracking and Counting Vehicles: While YOLOv8n excels in
accurately detecting and classifying vehicles, the Deep Sort algorithm takes it a step further
by introducing vehicle tracking and counting capabilities. Deep Sort combines the power of
deep learning and multi-object tracking to create a robust system that can handle complex
scenarios efficiently.

E) Accurate Tracking and ID Maintenance: Deep Sort employs sophisticated techniques to

associate the detected vehicles over consecutive frames, ensuring precise tracking. By
assigning unique IDs to each vehicle, it becomes possible to monitor individual vehicles'
movements, speeds, and behaviors accurately.

2
F) Real-Time Speed Estimation: Another significant aspect of the YOLOv8n and Deep Sort-
based system is real-time speed estimation. By utilizing the tracked vehicles' positions over
time, this system can estimate the speed at which each vehicle is traveling with impressive
accuracy.

1.1 Deep Learning

Deep learning is the branch of machine learning which is based on artificial neural
network architecture. An artificial neural network or ANN uses layers of interconnected
nodes called neurons that work together to process and learn from the input data.

In a fully connected Deep neural network, there is an input layer and one or more
hidden layers connected one after the other. Each neuron receives input from the previous
layer neurons or the input layer. The output of one neuron becomes the input to other
neurons in the next layer of the network, and this process continues until the final layer
produces the output of the network. The layers of the neural network transform the input
data through a series of nonlinear transformations, allowing the network to learn complex
representations of the input data.

Applications of Deep Learning:

The main applications of deep learning can be divided into computer vision, natural
language processing (NLP), and reinforcement learning.

Computer vision
In computer vision, Deep learning models can enable machines to identify and understand
visual data. Some of the main applications of deep learning in computer vision include:
 Object detection and recognition: Deep learning model can be used to identify
and locate objects within images and videos, making it possible for machines to
perform tasks such as self-driving cars, surveillance, and robotics.
 Image classification: Deep learning models can be used to classify images into
categories such as animals, plants, and buildings. This is used in applications
such as medical imaging, quality control, and image retrieval.

3
  Image segmentation: Deep learning models can be used for image
segmentation into different regions, making it possible to identify specific
features within images.

Object Detection

Object detection is the task of detecting instances of objects of a specific class within an
image or video. Basically, it locates the existence of objects in an image using a bounding
box and assigns the types or classes of the objects found. For instance, it takes an image as
input and generates one or more bounding boxes, each with the class label attached. These
algorithms are powerful enough to handle multi-class classification and localization and
objects with multiple occurrences.

Object detection is a combination of two tasks:

 Image classification
 Object localization

Image classification algorithms predict the type or class of an object in an image among a
predefined set of classes that the algorithm was trained for. Usually, input is an image with a
single object, such as a cat. Output is a class or label representing a particular object, often
with a probability of that prediction.

Object localization algorithms locate the presence of an object in the image and represent its
location with a bounding box. They take an image with one or more objects as input and
output the location of one or more bounding boxes using their position, height, and width.

4
Object Detection Methods

Generally, object detection methods can be classified as either neural network-based or non-
neural approaches. Also, some of them are rule-based, where the rule is predefined to match
specific objects. Non-neural approaches require defining features using some feature
engineering techniques and then using a method such as a support vector machine (SVM) to
do the classification.

Some of the non-neural methods are:

 Viola-Jones object detection method based on Haar features

 Scale-invariant feature transform (SIFT)
 Histogram of Oriented Gradients (HOG)
 Other methods based on template, shape, or color matching

On the other hand, neural network techniques can do end-to-end object detection without
explicitly defining features. They are far more accurate than non-neural based and are
typically built on convolutional neural networks (CNN).

Some of the neural network methods are:

 Region-Based Convolutional Neural Networks (R-CNN, Fast R-CNN, etc.)

 Single Shot Detector (SSD)
 Retina-Net
 You Only Look Once (YOLO)

Challenges in Object Detection

In object detection, the bounding boxes are always rectangular. As a result, if the object
contains the curvature part, it does not help determine its shape. In order to find precisely the
shape of the object, we should use some of the image segmentation techniques.

Some non-neural methods may not detect objects with high accuracy or may produce a large
number of false-positive detections. Although neural network methods are more accurate,
there are some drawbacks. For example, they require a large amount of annotated data for

5
training. Training is often expensive in time and space and, as a result, prolonged on standard
computers.

In order to solve these challenges, we can use the YOLO algorithm. Thanks to the transfer
learning capabilities, we would be able to use already pre-trained models or spend some time
fine-tuning models with our data. Furthermore, the YOLO algorithm is one of the most
popular methods for performing object detection in real-time because it achieves high
accuracy on most real-time processing tasks while maintaining a reasonable speed and frames
per second, even on devices accessible to almost everyone.

Object Detection Evaluation Metrics:

There are several metrics commonly used to evaluate the performance of object detection
models:

1. Mean Average Precision (mAP): The precision of a model is defined as the number
of true positives divided by the number of true positives plus false positives. The mAP metric
takes into account both the precision and recall of the model and is calculated as the mean of
the average precision for each class. A higher mAP value indicates a better performance of
the model.

2. Average Precision (AP): It is a measure of the precision of the model at different

recall levels. The precision is defined as the number of true positives divided by the number
of true positives plus false positives. The recall is defined as the number of true positives
divided by the number of true positives plus false negatives. AP is calculated as the area
under the precision-recall curve. A higher AP value indicates a better performance of the
model.

3. Intersection over Union (IoU): IoU is a measure of the overlap between the
predicted bounding box and the ground-truth bounding box. It is calculated by taking the
intersection areas of the two boxes and dividing it by the area of their union. A higher IoU
value indicates a better match between the predicted and ground-truth bounding boxes.

6
There are also several metrics that evaluate the accuracy too. For example, True Positive Rate
(TPR), False Positive Rate (FPR), F1-score, and Log Average Miss Rate (MR). In addition to
these metrics, object detection models can also be evaluated based on their computational
efficiency.

You Only Look Once (YOLO)

You Only Look Once (YOLO) is one of the most popular model architectures and object
detection algorithms. It uses one of the best neural network architectures to produce high
accuracy and overall processing speed, which is the main reason for its popularity. If we
search Google for object detection algorithms, the first result will be related to the YOLO
model.

YOLO algorithm aims to predict a class of an object and the bounding box that defines the
object location on the input image. It recognizes each bounding box using four numbers:

 Center of the bounding box (bx, by)

 Width of the box (bw)
 Height of the box (bh)

In addition to that, YOLO predicts the corresponding number c for the predicted class as well
as the probability of the prediction (pc)

7
YOLO has been developed in several versions, such as YOLOv1, YOLOv2, YOLOv3,
YOLOv4, YOLOv5, YOLOv6, and YOLOv7. Each version has been built on top of the
previous version with enhanced features such as improved accuracy, faster processing, and
better handling of small objects.

YOLO is widely used in various applications such as self-driving cars and surveillance
systems. It is also widely used for real-time object detection tasks like in real-time video
analytics and real-time video surveillance.

Fig 1: A timeline of YOLO versions.

YOLO Strengths and Limitations

Although YOLO is a fast object detector, it has some limitations. YOLO has a more
significant localization error than state-of-the-art methods like Fast R-CNN. This limitation
can be attributed to three major causes:
8
 1. YOLO can only detect a maximum of two objects of the same class in the grid cell,
which limits its ability to predict nearby things.
2. YOLO has difficulty predicting objects with aspect ratios that were not present in the
training data.
3. YOLO learns from coarse object features due to the down-sampling process.

YOLOV2
YOLO v2, also known as YOLO 9000, is an improved version of the original
YOLO object detection algorithm. It builds upon the concepts and architecture of
YOLO, but addresses some of the limitations of the original version.
One of the main differences between YOLO v2 and the original YOLO is the use
of anchor boxes. In YOLO v2, ‌CNN predicts not only the bounding box
coordinates but also the anchor boxes. Anchor boxes are pre-defined boxes of
different aspect ratios and scales, which are used to match the predicted bounding
boxes with the actual objects in the image. This allows YOLO v2 to handle objects
of different shapes and sizes better.
Another key difference is the use of a multi-scale approach. In YOLO v2, the input
image is fed through ‌CNN at multiple scales, which allows the model to detect
objects at different sizes. This is achieved by using a feature pyramid network
(FPN), which allows the model to extract features at different scales from the same
image.
Additionally, YOLO v2 uses a different loss function than the original YOLO,
called the sum-squared error (SSE) loss function. The SSE loss function is more
robust and helps the model to converge faster.
In terms of architecture, YOLO v2 uses a slightly deeper CNN than YOLO, which
allows it to extract more powerful features from the image. The CNN is followed
by several fully connected layers, which predict ‌class probabilities and bounding
box coordinates.

9
 Fig: mAP-and-FPS-of-YOLO-v2

YOLO v3
YOLO v3 is the third version of the YOLO object detection algorithm. The first
difference between YOLO v3 and previous versions is the use of multiple scales in
the input image. YOLO v3 uses a technique called "feature pyramid network"
(FPN) to extract features from the image at different scales. This allows the model
to detect objects of different sizes in the image.
Another important difference is the use of anchor boxes. In YOLO v3, anchor
boxes are used to match the predicted bounding boxes with the actual objects in the
image. Anchor boxes are pre-defined boxes of different aspect ratios and scales,
and the model predicts the offset of the anchor boxes relative to the bounding
boxes. This helps the model to handle objects of different shapes and sizes better.
In terms of architecture, YOLO v3 is built on a deep convolutional neural network
(CNN) that is composed of many layers of filters. The CNN is followed by several
fully connected layers, which predict ‌class probabilities and bounding box
coordinates.
YOLO v3 also uses a different loss function than previous versions. It uses a
combination of classification loss and localization loss, which allows the model to
learn both the class probabilities and the bounding box coordinates.

10
 YOLO v3 speed and mAP compared with other detectors

YOLO v4
Now, let’s take a look at one of the most important versions of YOLO. A key
distinction between YOLO v4 and previous versions is using a more advanced
neural network architecture. YOLO v4 uses a technique called "Spatial Pyramid
Processing" (SPP) to extract features from the image at different scales and
resolutions. This allows the model to detect objects of different sizes in the image.

Additionally, YOLO v4 also uses a technique called "Cross-stage partial

connection" (CSP) to improve the model's accuracy. It uses a combination of
multiple models with different architectures and scales and combines their
predictions to achieve better accuracy.

11
 Fig: YOLO v4 accuracy and FPS compared with other detectors

YOLO v5
YOLO v5 was introduced in 2020 with a key difference from the previous
versions, which is the use of a more efficient neural network architecture called
EfficientDet, which is based on the EfficientNet architecture. EfficientDet is a
family of image classification models that have achieved state-of-the-art
performance on a number of benchmark datasets. The EfficientDet architecture is
designed to be efficient in terms of computation and memory usage while also
achieving high accuracy.

Another important difference is the use of anchor-free detection, which eliminates

the need for anchor boxes used in previous versions of YOLO. Instead of anchor
boxes, YOLO v5 uses a single convolutional layer to predict the bounding box
coordinates directly, which allows the model to be more flexible and adaptable to
different object shapes and sizes.
YOLO v5 also uses a technique called "Cross mini-batch normalization" (CmBN)
to improve the model's accuracy. CmBN is a variant of the standard batch
normalization technique that is used to normalize the activations of the neural
network.

12
Regarding training, YOLO v5 uses transfer learning, which allows it to be pre-
trained on a large dataset and then fine-tuned on a smaller dataset. This allows the
model to learn from a wide range of data and generalize better to new data.

Fig: Speed of YOLO v5 compared with EfficientDet [repository]

YOLO v6
A notable contrast between YOLO v6 and previous versions is the use of a more
efficient and lightweight neural network architecture; this allows YOLO v6 to run
faster and with fewer computational resources. The architecture of YOLO v6 is
based on the "EfficientNet-Lite" family, which is a set of lightweight models that
can be run on various devices with limited computational resources.

YOLO v6 also incorporates data augmentation techniques to improve the

robustness and generalization of the model. This is done by applying random
transformations to the input images during training, such as rotation, scaling, and
flipping.

13
 Fig: YOLO v6 accuracy and speed
YOLO v7
YOLO v7, on the most recent stable iterations of the YOLO algorithm. It boasts a
number of enhancements compared to ‌previous versions. A key enhancement is the
implementation of anchor boxes. These anchor boxes, which come in various
aspect ratios, are utilized to identify objects of various shapes. The use of nine
anchor boxes in YOLO v7 enables it to detect a wider range of object shapes and
sizes, leading to a decrease in false positives.
In YOLO v7, a new loss function called "focal loss" is implemented to enhance
performance. Unlike the standard cross-entropy loss function used in previous
versions of YOLO, focal loss addresses the difficulty in detecting small objects by
adjusting the weight of the loss on well-classified examples and placing more
emphasis on challenging examples to detect.

14
 YOLOv7-speed-and-accuracy

YOLOv8

YOLOv8 is the latest family of YOLO-based object detection models from Ultralytics that
provides state-of-the-art performance.

Leveraging the previous YOLO versions, the YOLOv8 model is faster and more accurate
while providing a unified framework for training models for performing:

 Object Detection
 Object Tracking
 Instance Segmentation
 Image Classification

15
 Figure 1: YOLOv8 architecture

Ultralytics has released a complete repository for YOLO Models. Also, Ultralytics will
release a paper on Arxiv comparing YOLOv8 with other state-of-the-art vision models. We
will not discuss YOLOv8 here; read our detailed article about YOLOv8 to learn more about
YOLOv8 and its applications. In this article, we will explore YOLOv8 object tracking. And
how we can use YOLOv8 and OpenCV for object tracking and counting.

Using YOLOv8 for Object Detection

In this article, we will explore YOLOv8 object tracking and counting. It’s important to note
that for effective object tracking, we require inputs from an object detection system, in this
case, YOLOv8.Object detection is a task where we localize and classify objects in an image
or sequence of video frames. You can read our article about moving object detection using
OpenCV, where we discussed how to detect objects just using OpenCV.

16
OpenCV for YOLOv8 Object Tracking

OpenCV is our most extensive open-sourced library for computer vision, containing almost
every possible image-processing algorithm. Leveraging OpenCV for YOLOv8 Object
Tracking combines the advanced detection capabilities of YOLOv8 with the robust features
of the OpenCV library, offering an innovative solution for sophisticated real-time object
tracking. Let’s explore this unique combination in more detail.

OpenCV primarily provides eight different trackers available in OpenCV 4.2 — BOOSTING,
MIL, KCF, TLD, MEDIANFLOW, GOTURN, MOSSE, and CSRT out of the box. A
majority of the open-source trackers use OpenCV for most of the visualization and image
processing works. Learn more about OpenCV Object Tracking in our detailed article Object
Tracking using OpenCV.

Object Tracking

YOLOv8 Object Tracking is an extended part of object detection where we identify the
location and class of objects within the frame and maintain a unique ID for each detected
object in subsequent video frames.

17
Method Based Classification

Application Based Classification

We can classify Object Tracking into two main categories based on the type and functionality
of available trackers.

18
Single Object Tracker (SOT)

In this class of trackers, the first frame is marked using a rectangle to indicate the location of
the object we want to track. The object is then tracked in subsequent frames using the
tracking algorithm. In most real-life applications, these trackers are used with an object
detector. OpenCV has built-in trackers for you, including KCF, TLD, and GOTURN. You
can check our detailed article about Single Object Trackers for more information.

Multiple Object Tracker (MOT)

With a fast object detector, detecting multiple objects and then running a tracker to track
multiple objects in a single or consecutive frame makes sense.

Various advanced Multi-Object Tracking (MOT) systems exist, such as Deep

SORT, FaitMOT, Byte Track, Bot SORT, etc. These systems employ sophisticated
algorithms to track multiple objects accurately and efficiently in video sequences.

19
Multiple Object Tracking (MOT) algorithms is tracking-by-detection, where detections
(bounding boxes identifying targets in video frames) guide the tracking process. These
detections are associated with maintaining consistent IDs for the same targets across frames.
This makes MOT primarily an assignment problem. Thanks to modern detection frameworks
ensuring high-quality detections, most MOT methods focus on enhancing Data
association rather than detection. Many MOT datasets provide pre-determined detections,
allowing algorithms to bypass detection and focus solely on comparing association quality.

1.4.1 Installation and Usage

To implement Ultralytics provides various installation methods including pip, conda,

and Docker. Install YOLOv8 via the ultralytics pip package, that can be installed using PyPl:

pip install ultralytics

1.4.2 Steps in Fuzzy C-Means Clustering

The process flow of fuzzy c-means is enumerated below:

1. Assume a fixed number of clusters k.

2. Initialization: Randomly initialize the k-means μk associated with the clusters and

compute the probability that each data point xi is a member of a given cluster k,

P(point xi has label k|xi, k).

3. Iteration: Recalculate the centroid of the cluster as the weighted centroid given the

probabilities of membership of all data points xi:

20
 4. Termination: Iterate until convergence or until a user-specified number of iterations has

been reached (the iteration may be trapped at some local maxima or minima).

Fig 1.4.2 Steps in Fuzzy C-Means Clustering

1.4.3 Advantages of Fuzzy C means

 Works better than the standard hard clustering algorithm, i.e, k-means algorithm

especially when we have to deal with overlapping data points.

 Unlike k-means where data point must exclusively belong to one cluster center here

data point is assigned membership to each cluster center as a result of which data

point may belong to more than one cluster center.

1.4.4 Disadvantages Of Fuzzy C means

 Inappropriate specification of the number of clusters.

21
  With lower value of β we get the better result but at the expense of more number of

iteration.

 Euclidean distance measures can unequally weight underlying factors.

1.5 TECHNOLOGY USED

1.5.1 Introduction to Python

Python is developed by Guido van Rossum. Guido van Rossum started implementing

Python in 1989. Python is a facile programming language so even if you are new to

programming, you can learn python without facing any issues. Python is a general-purpose

programming language that is becoming ever more popular for data science. Companies

worldwide are using Python to harvest insights from their data and gain a competitive edge.

Python specifically for data science. To store and manipulate data, and helpful data science

tools to begin conducting your own analyses.

Python is an interpreted, high-level and general-purpose programming language. It is

dynamically typed and collected. Python is an interpreted language and not a compiled one,

although compilation is a step. Python code, written in .py file is first compiled to what is

called byte code which is stored with a .pyc or. pyo format. Instead of translating source code

to machine code like C++, Python code it translated to byte code. This byte code is a low-

level set of instructions that can be executed by an interpreter. One popular advantage of

interpreted languages is that they are platform-independent. As long as the Python byte code

and the Virtual Machine have the same version, Python byte code can be executed on any

platform (Windows, MacOS, etc.). Dynamic typing is another advantage. In static typed

languages like C++, you have to declare the variable type and any discrepancy like adding a

string and an integer is checked during compile time. In older programming languages,

22
memory allocation was quite manual. Many times, when you use variables that are no longer

in use or referenced anywhere else in the program, they need to be cleaned from the memory.

Garbage Collector does that for you.

Fig. 1.5 Working of Python

1.5.2 PyCharm

PyCharm basically provide an interactive computational environment for developing

python-based Data science applications. A PyCharm app is a server –client application that

allows running and editing notebook document via a web browser. PyCharm combines live

code, graphics, visualizations, and text in shareable notebooks that run in a web-browser, the

notebook itself could be hosted on your local machine or on a remote server. It is an open-

source, web-based interactive environment which allows you to create and share documents

that contain live code, mathematical equation, graphics, maps, plots, visualizations and

narrative text.

CHAPTER 2

LITERATURE SURVEY

R. Kejriwal, R. H J, A. Arora and Mohana et.al., "Vehicle Detection and Counting using
Deep Learning based YOLO and Deep SORT Algorithm for Urban Traffic Management
23
System," The method for determining how congested a route is for Intelligent Transportation
Systems (ITS) is the main emphasis of this work. The primary issues addressed by the current
traffic systems are speeding and signal infractions, but they frequently overlook the volume
of traffic. The suggested method counts and tracks automobiles using footage from mobile
and CCTV cameras in smart cities. A YOLO-based deep learning model is used to identify,
count, and classify automobiles in the video. Then, several vehicles are tracked in the video
frames using the Deep SORT algorithm. The model has been carefully trained for the
conditions of Indian roads. According to the review, the model's categorization and counting
accuracy is mediocre. In order to improve infrastructure, this model can assist in determining
how congested the roadways are. It plays a significant role in enhancing the intelligence and
effectiveness of traffic surveillance in smart cities [4].

Zhaoming Zhou, Hui Li.et.al., “Vehicle Object Detection Based on Deep Learning ” The
increased use of autos and the corresponding growth in activity mishaps necessitate strong
vehicle recognition. Specifically, it focuses on using deep learning, with an emphasis on the
SSD (Single Shot Multi-Box Locator) computation. Finding a reliable calculation with a high
acknowledgment rate is the goal. The SSD computation and the central convolutional neural
network (CNN) theory are examined in detail in this research. At that point, it tests the
method by determining whether the vehicles are moving or stationary based on a video and
images of stationary vehicles. Picture recognition is used to assess the accuracy and
practicality of the suggested method, which provides a practical solution for vehicle
recognition with a high degree of solidity and precision [11].

24
Rouf, Md Abdur, Wu, Qing, Yu, Xiaoyu, Iwahori, Yuji, Wu, Haibin, Wang, Aili, et.al.,
“Real-time Vehicle Detection, Tracking and Counting System Based on YOLOv7” The
importance of a YOLOv7-based real-time vehicle monitoring system for highway
applications is highlighted by this study. The system's primary goal is to reduce traffic
congestion, violations, and road incidents by improving traffic management, planning, and
accident prevention. It provides real-time traffic flow statistics by recognizing, tracking, and
keeping an eye on a variety of vehicle types on urban roads and highways. The system's
essential phases include classification, tracking, monitoring, and vehicle detection. A revised
version of YOLOv7 is proposed to overcome some of its drawbacks, including low feature
extraction and high missed detection rates. The goal of this upgraded system is to provide
real-time vehicle tracking, providing information on traffic dynamics and assisting in well-
informed traffic management decision-making. The study essentially presents a YOLOv7-
based architecture for accurate real-time vehicle identification, tracking, and counting,
enhancing our comprehension of traffic patterns and aiding in traffic management decisions
[12].

Jian-Da Wu; Bo-Yuan Chen; Wen-Jye Shyr; Fan-Yu Shih.et.al., “Vehicle Classification and
Counting System Using YOLO Object Detection Technology” By integrating digital traffic
monitoring technology into an intelligent transportation system, this initiative seeks to
improve urban transportation. Traffic data is collected by elevated surveillance cameras, and
various vehicle kinds are identified through the use of YOLO object identification
technology. The study focuses on particular urban settings, including T-shaped intersections,
cross-type intersections, and straight-line bidirectional roads and uses deep convolutional
neural networks and visual processing for object detection training. In order to accurately
count the number of vehicles, the system uses object tracking to draw a line in the direction
of the pathways taken by the vehicles and tracks the center coordinate as objects cross this
line [13].

Lin CJ, Jeng SY, Lioa HW, et.al., A Real-Time Vehicle Counting, Speed Estimation, and
Classification System Based on Virtual Detection Zone and YOLO” The Open Works and
Open Lodging Service (PUPR) developed a web application utilizing the deep learning model
YOLOv4 in order to increase the efficiency of their manual traffic surveys. Based on Bina
Marga classifications, this model can recognize and classify cars automatically. Using video
25
data from traffic records, the app generates a car count report. During testing, a small
collection of 1000 images were used for training on the Google Colab platform, and the
YOLOv4 model, along with the Deep SORT technique and a bespoke dataset, obtained a
respectable accuracy [15].

CHAPTER 3

SYSTEM STUDY

3.1 EXISTING SYSTEM

In existing System, Movie recommender system was done by the K-Means and

KNN Algorithm to give the recommended system for the user’s predicts the user’s preference

of a movie on the basis of different parameters. The recommender system works on the

concept that people are having common preference or choice.

These users will influence on each other’s opinions. This process optimizes the

process and having lower RMSE.

3.1.1 Issues in the system:

 Doesn't work well with a large dataset

 Doesn't work well with a high number of dimensions

 Sensitive to outliers and missing values

 It is not suitable to identify clusters with non-convex shapes.

3.1.2 Solutions:

 Ditch User-Based Collaborative Filtering Model.

 A Gold Standard Similarity Computation Technique.

 Boost Your Algorithm Using Model Size.

26
3.2 PROPOSED SYSTEM

In response to the growing significance of highways and the increasing role of intelligent
vehicle detection, counting, and speed estimation, this paper tackles the persistent challenge
posed by the diverse sizes of vehicles. The accurate detection of vehicles becomes
particularly challenging when dealing with variations in size. The research suggests a vision-
based method for vehicle identification, counting, and speed calculation in order to get
around this problem. This methodology seeks to improve the accuracy of these important jobs
by utilizing cutting-edge computer vision techniques, providing a viable solution for the
changing environment of intelligent transportation systems on roads.

Fig.1: Block diagram of vehicle detection, counting and speed estimation.

Fig.1 displays the block diagram of proposed method. In the field of vehicle object
recognition, deep convolutional networks (CNNs) have shown great success by showcasing
their exceptional ability to learn complex picture characteristics and perform a range of tasks,
such as bounding box regression and classification. Generally speaking, there are two main
types of detection techniques. The first, a two-step process, uses a convolutional neural
network for object categorization after creating a candidate box for the object using several
techniques. On the other hand, the second method, which is a one-stage approach, streamlines
the procedure by simply transforming the bounding box positioning problem of the item into

27
a regression problem. This eliminates the requirement of creating a candidate box. Using both
methods, CNNs can handle challenging jobs like vehicle detection [10].

This project is aimed at implementing the YOLO-V8-based Deep SORT model for detecting
and tracking vehicles in video sequences in real-time. The addition of the Deep SORT
algorithm enhances the tracking of actual vehicles identified by YOLOv8 in video frames.
The video is segmented into frames, serving as input for YOLOv8 to detect vehicles.
Detected frames are then analyzed by the Deep SORT algorithm for vehicle tracking. If a
vehicle is tracked, Deep SORT overlays a bounding box and increments the tracking counts.
The model is trained using COCO datasets.

Fig.2: The methodology of proposed system.

CHAPTER 4

IMPLEMENTATION

4.1 MODULE

Dataset

Data Pre-processing and Frame Separation

Resize images

YOLO V8n

28
Deep Sort

4.2 MODULES DESCRIPION

4.2.1. DATASET

The COCO (Common Objects in Context) dataset is a large-scale object detection,

segmentation, and captioning dataset. It is designed to encourage research on a wide variety
of object categories and is commonly used for benchmarking computer vision models. It is an
essential dataset for researchers and developers working on object detection, segmentation,
and pose estimation tasks.

A) Key Features

1. COCO contains 330K images, with 200K images having annotations for object
detection, segmentation, and captioning tasks.

2. The dataset comprises 80 object categories, including common objects like cars,
bicycles, and animals, as well as more specific categories such as umbrellas,
handbags, and sports equipment.
3. Annotations include object bounding boxes, segmentation masks, and captions for
each image.

4. COCO provides standardized evaluation metrics like mean Average Precision (mAP)
for object detection, and mean Average Recall (mAR) for segmentation tasks, making
it suitable for comparing model performance.

Dataset Structure

The COCO dataset is split into three subsets:

1. Train : This subset contains 118K images for training object detection,
segmentation, and captioning models.

2. Val : This subset has 5K images used for validation purposes during model
training.

29
 3. Test : This subset consists of 20K images used for testing and
benchmarking the trained models. Ground truth annotations for this subset are
not publicly available, and the results are submitted to the COCO evaluation
server for performance evaluation.

To train a YOLOv8n model on the COCO dataset for 100 epochs with an image size of
640, you can use the following code snippets. For a comprehensive list of available
arguments, refer to the model Training page.

FIG: LOAD A DATASET

4.2.2 DATA PRE-PROCESSING AND FRAME SEPARATION

Digital image processing is the use of computer algorithms to perform image

processing on digital images. As a subfield of digital signal processing, digital image
processing has many advantages over analogue image processing. It allows a much wider
range of algorithms to be applied to the input data

— the aim of digital image processing is to improve the image data (features) by suppressing
unwanted distortions and/or enhancement of some important image features so that our AI-
Computer Vision models can benefit from this improved data to work on. To train a network
and make predictions on new data, our images must match the input size of the network. If
they need to adjust the size of images to match the network, then they can rescale or crop data
to the required size.

they can effectively increase the amount of training data by applying randomized
augmentation to data. Augmentation also enables to train networks to be invariant to
distortions in image data. For example, they can add randomized rotations to input images so

30
that a network is invariant to the presence of rotation in input images. An augmented Image
Datastore provides a convenient way to apply a limited set of augmentations to 2-D images
for classification problems. They can store image data as a numeric array, an Image Datastore
object, or a table. An Image Datastore enables to import data in batches from image
collections that are too large to fit in memory. they can use an augmented image datastore or
a resized 4-D array for training, prediction, and classification. They can use a resized 3-D
array for prediction and classification only.

There are two ways to resize image data to match the input size of a network. Rescaling
multiplies the height and width of the image by a scaling factor. If the scaling factor is not
identical in the vertical and horizontal directions, then rescaling changes the spatial extents of
the pixels and the aspect ratio.

Cropping extracts a subregion of the image and preserves the spatial extent of each pixel.
They can crop images from the center or from random positions in the image. An image is
nothing more than a two-dimensional array of numbers (or pixels) ranging between 0 and
255. It is defined by the mathematical function f (x, y) where x and y are the two co-ordinates
horizontally and vertically.

31
3.3 YOLO v8n

YOLOv8n is built upon a deep convolutional neural network (CNN) architecture, resembling
its earlier versions. The functioning of YOLOv8n involves a systematic process. It starts by
dividing the input image into a grid of cells. In each cell, YOLOv8n makes predictions for
bounding boxes, accompanied by corresponding class probabilities. The model's output
provides multiple bounding boxes for potential objects present in the image. To refine the
results and ensure accuracy, YOLOv8n employs a non-maximum suppression (NMS)
algorithm. This algorithm filters out overlapping bounding boxes, selecting the most probable
ones for each object in the image. In essence, YOLOv8n efficiently processes images through
this grid-based approach, enabling accurate and fast object detection while minimizing
redundancy in the results.

The industry standard benchmark used to evaluate object identification model performance is
the Common Objects in Context (COCO) dataset. Important parameters like FPS (frames per
second) for inference speed and mAP (mean average precision) for accuracy are taken into
account while evaluating COCO. It is important to make sure that comparisons between
models on the COCO dataset are fair by looking at mAP values and FPS measurements at
comparable inference speeds. When compared to models with identical inference latencies,
YOLOv8, a variation of the You Only Look Once (YOLO) object identification algorithm,
has shown state-of-the-art accuracy on the COCO dataset as of this writing. This
demonstrates how well YOLOv8 performs in object detection tasks by achieving excellent
precision at competitive inference times.

Variations of Yolo V8: There are three variations of Yolo V8: YOLOv8-n, YOLOv8-L, and
YOLOv8-X. The backbone network's size is the primary distinction between the variations.
The backbone network of YOLOv8 is the smallest, and that of YOLOv8-X is the largest.
Although YOLOv8-X has a larger backbone network than YOLOv8n and YOLOv8-L, it is
slower overall. This is due to the increased accuracy.

32
 Fig.3: Detection of vehicles using YOLOV8.

3.4 Deep Sort

The Deep SORT algorithm is initialized to enable robust object tracking. The program
manages a deque (DQ) list, appending the bottom coordinates of detected objects and
removing them as they exit the frame. Bounding box coordinates are assigned unique IDs for
each object, facilitating tracking, and displays the tracked objects along with their IDs.
Additionally, vehicle counting is implemented using a get direction function, determining the
direction of objects entering or leaving the frame. The code demonstrates the vehicle
counting process through a downloadable and runnable file. Objects moving towards the
north are considered to be entering, while those moving towards the south are deemed to be
leaving [4].

A vehicle counts the number of vehicles entering and exiting, and has a place set aside to
display the counts. When a vehicle crosses a predetermined line, the implementation
increments the count to ensure accurate counting. The direction of the vehicle decides
whether the count represents vehicles entering or leaving. This all-inclusive method enables
accurate vehicle counting in the designated directions as well as efficient object tracking [10].

3.5 Speed Estimation

A real-time speed estimation system leveraging YOLOv8 and Deep SORT combines the
speed and accuracy of YOLO's vehicles detection with the robust vehicles tracking
capabilities of Deep SORT. In this integrated framework, YOLOv8 is employed to detect and
identify vehicles in real-time video frames, while Deep SORT is utilized for continuous
33
tracking of the identified vehicles across consecutive frames. By measuring the distance
traveled by each vehicle and recording the corresponding time intervals between frames, the
system calculates real-time speed estimates. This approach provides a comprehensive
solution for accurate and efficient speed estimation in dynamic environments, making it
applicable for traffic monitoring, intelligent transportation systems, and other scenarios
requiring precise real-time speed data.

Fig .4: Speed Estimation Code.

Fig .5:Vehicle Speed Estimation

CHAPTER 6

RESULT AND CONCLUSION

6.1 RESULT AND DISCUSSION

To implement this project, we have designed following modules.

1. Generate & Load YOLOv8-DeepSort Model: using this module we will

generate and load YOLOV8-Deep-Sort model.

34
 2. Upload Video & Detect Car & Truck: using this module we will upload test video

and then apply YOLOV8 to detect vehicle and this detected vehicle frame will be further

analyse by Deep-Sort to track real vehicles.

3. The pre-trained YOLOv8 model is obtained by training the model on the COCO

dataset.

The YOLOv8n and Deep SORT algorithms have been used for vehicle classification and
counting using the COCO dataset and video clips. Based on a predetermined boundary line,
the system counts the number of vehicles entering and exiting the frame and classifies them
using speed estimate. YOLOv8n is especially useful for real-time surveillance applications. It
can be used to record videos to evaluate the flow of vehicles over predetermined time
intervals or to monitor traffic on highways. The identification of regions prone to congestion
is made possible by the considerable potential this data offers for traffic management.
YOLOv8n is the recommended option for situations where efficiency and precision are
essential since it performs better than earlier generations in terms of both speed and accuracy.

Furthermore, the Deep SORT-based vehicle tracking algorithm has remarkable precision in
identifying the nonlinear motion of automobiles. The suggested technique improves tracking
accuracy by utilizing YOLOv8n with Dark-net, an open-source neural network framework,
for vehicle localization and identification. This all-encompassing method has the potential to
enhance data-driven traffic management strategies and offer highly accurate vehicle tracking
for advanced traffic monitoring systems.

6.2 CONCLUSION

In conclusion, the vehicle detection and tracking, classifying, and speed estimating method
presented uses Ultralytics library with Deep-SORT algorithm based on YOLOv8n model. It
can be use in Realtime surveillance camera in the highway or recording video to evaluate the
number of vehicles pass by according to what time it started recorded to last recorded. This
data then can be used for traffic management by implementing answer if the place proven a
lot of congestion or not. It is the best to use YOLOv8 model than previous model YOLOv7 if
the system wants the highest accuracy with acceptable speed. If the system wants the best
accuracy with the highest speed as possible because limitation in hardware or to process it in
real-time, it is recommended to use YOLOV8 model which it can achieve higher accuracy.
35
This system can be improved to be more adaptable for vehicle detection if using several
suggestion ideas. A vehicle tracking algorithm based on the framework suggested in Deep-
SORT which is capable of tracking the nonlinear motion of vehicles with a high level of
accuracy. The proposed algorithm utilizes YOLOv8 with Darknet, an open-source neural
network framework, for vehicle localization and identification. The number of detection
errors was minimized by optimizing the training of the detector through hyperparameter
optimization and data augmentation.

we suggested a YOLOv8n-based real-time vehicle tracking, detection, and classification

system. The suggested system can be used to monitor traffic flow on highways and is capable
of reliably detecting, tracking, classifying, and speed estimating various sorts of vehicles in
real-time. Additionally, we have created the predict.py model, which is capable of object
detection, tracking, object counting, and real-time video. Transportation authorities can make
better judgments to increase the effectiveness and safety of the transportation system by using
the system's insightful data on highway traffic flow.

Future development will focus on improving accuracy, which will include features such as
lane recognition for over-speed alarms.

REFERENCES

1. Mohammadnazar V, Arvin R, Khattak AJ, classifying travelers' driving style using basic
safety messages generated by connected vehicles: Application of unsupervised machine
learning,” Transp Res Part C Emerg Technol. 2021; 122. doi: 10.1016/j.trc.2020.102917.

2. Wang Y, Ban X, Wang H, et al. Detection and Classification of Moving Vehicle From
Video Using Multiple Spatio Temporal Features, In IEEE Acce. 2019; 7: 80287-99p. doi:
10.1109/ACCESS.2019.2923199.

3. Kusumah AP, Djayusman D, Setiadi GR, et al. Counting Various Vehicles Using
YOLOv4 and DeepSORT, JIAE. 3(1) :1-6p. doi:10.51662/jiae.v3i1.68.

36
4. Kejriwal R, Ritika HJ, Arora A, et al (Eds). Vehicle Detection and Counting using Deep
Learning based YOLO and Deep SORT Algorithm for Urban Traffic Management System.
First International Conference on Electrical, Electronics, Information and Communication
Technologies (ICEEICT); 2022; Trichy, India. 2022. 1-6p.

5. Gomaa A, Abdelwahab MM , Abo-Zahhad M, et al . Robust Vehicle Detection and

Counting Algorithm Employing a Convolution Neural Network and Optical Flow, Sensors.
2019, 19(20), 4588.https://doi.org/10.3390/s19204588.

6. Sonnleitner E, Barth O, Palmanshofer A, et al.Traffic Measurement and Congestion

Detection Based on Real-Time Highway Video Data, Appl. Sci.2020,10(18) 6270.
https://doi.org/10.3390/app10186270.

7. Vishal M, Yaw AG, Object Detection and Tracking Algorithms for Vehicle Counting: A
Comparative Analysis, J. Big Data Anal. Transp.2020. https://doi.org/10.1007/s42421-020-
00025-w

8. Meghana RK, Apoorva S, Chitkara Y(Eds). Background-modelling techniques for

foreground detection and Tracking using Gaussian Mixture Model.3rd International
Conference on Computing Methodologies and Communication (ICCMC); 2019;1129-1134p.
DOI:10.1109/ICCMC.2019.8819825

9. Kumar CB, Punitha R, Mohana (Eds). Performance Analysis of Object Detection

Algorithm for Intelligent Traffic Surveillance System. Second International Conference on
Inventive Research in Computing Applications (ICIRCA); 2020;573-579P.
DOI:10.1109/ICIRCA48905.2020.9182793.

10. Lakshmi RA, Aishwarya CH, Sahithi1 A, et al. Real-time Vehicle Detection and Tracking
using YOLO-based Deep Sort Model: A Computer Vision Application for Traffic
Surveillance, Turk. J. Comput. Math. Educ.2023, 14 (01),255- 64p.

11. Zhou Z, Li H, Vehicle Object Detection Based on Deep Learning. J Sci Educ Technol,
2023,5(1), 38-45P.

37
12. Rouf Md A, Wu Q, Yu X, Yuji I, et al (Eds). Real-time Vehicle Detection, Tracking and
Counting System Based on YOLOv7. the international symposium on computer science,
computer engineering and educational technologies (ISCSET); 2023; doi:10.14464/ess.
v10i7.598.

13 Wu JD, Chen BY, Shyr WJ, et al. Vehicle Classification and Counting System Using
YOLO Object Detection Technology. Trait. du Signal, 2021, 38(40), 1087p.

14. Shine Rajesh D, Vaibhavi T, Sreeja T, et al. computer vision application: vehicle
counting and classification system from real time videos. Turk. J. Comput. Math. Educ,
2023,14(03),79- 89p.

15. Lin CJ, Jeng SY, Lioa HW, A Real-Time Vehicle Counting, Speed Estimation, and
Classification System Based on Virtual Detection Zone and YOLO. Math. Probl. Eng.,2021,
Article ID 1577614, 10p, https://doi.org/10.1155/2021/1577614.

38