Raj Kumar Goel Institute of Technology Ghaziabad

ISO 9001:2015 Certified

5th KM. STONE, DELHI-MEERUT ROAD, GHAZIABAD (U.P)-201003
Department of Computer Science & Engineering

Project Progress Report

1. Course : Bachelor of Technology

2. Semester : 7th
3. Branch : Computer Science & Engineering
4. Project Title : AI-Object Detection Using TensorFlow
5. Details of Students:

S. No. Roll No. Name Role as Signature

1 220033010900 Arjun Singhal Team Leader
5
2 210033010024 Tushar Tyagi Designer
1

3 210033010024 Tushar Report

0

6. SUPERVISOR:

Ms. Varsha Tyagi

Remarks from Project Supervisor:

………………………………………………………………………………

………………………………………………………………………………
2
………………………………………………………………………………

3
 SYNOPSIS

AI object detection is a critical task in computer vision that involves identifying and
classifying objects within images or video streams. It not only locates objects but
also labels them with their respective classes, making it vital for applications across
various industries. TensorFlow, an open-source machine learning framework devel-
oped by Google, offers a powerful environment for building, training, and deploying
object detection models. By leveraging pre-trained models and providing tools for
model optimization and real-time processing, TensorFlow simplifies the creation of
sophisticated object detection systems.

The object detection process typically involves several key stages. First, images are
preprocessed by resizing and normalizing them to meet the input requirements of the
model. Then, feature extraction takes place, often using convolutional neural net-
works (CNNs), which analyze the images for patterns and structures. Afterward, po-
tential object locations, or bounding boxes, are proposed using region proposal algo-
rithms like Region Proposal Networks (RPNs). Finally, the model classifies the de-
tected objects and refines their bounding boxes to improve accuracy.

The TensorFlow Object Detection API plays a central role in simplifying this
process. This API provides access to a range of pre-trained models such as Faster R-
CNN, SSD (Single Shot MultiBox Detector), and YOLO (You Only Look Once),
enabling rapid prototyping and deployment. TensorFlow’s transfer learning capabili-
ties allow users to fine-tune these pre-trained models on custom datasets, enhancing
their ability to recognize objects specific to different domains. The API also supports
TensorFlow Lite for deploying models on mobile and edge devices, and Tensor-
Flow.js for running object detection models directly within web browsers using
JavaScript.

Training object detection models with TensorFlow involves preparing a labeled

dataset with annotated bounding boxes and class labels. This data is used to train the
model, which learns to recognize patterns in the images that correspond to specific
objects. During training, the model attempts to minimize the classification and local-
ization loss functions, improving its accuracy in predicting both object labels and the
precise location of objects in images.

4
After training, the model’s performance is evaluated using metrics such as mean Av-
erage Precision (mAP), Intersection over Union (IoU), and precision-recall curves.
These metrics help assess how well the model can identify and localize objects
within test data. TensorFlow also offers tools for optimizing models to make them
more efficient, such as quantization, pruning, and support for TensorFlow Lite,
which reduces model size and improves inference speed for deployment on resource-
constrained devices.

Object detection has numerous real-world applications. In autonomous vehicles, ob-

ject detection is crucial for identifying pedestrians, other vehicles, traffic signs, and
obstacles in real time. In surveillance systems, it enables automated tracking of peo-
ple and objects in security footage. In retail and manufacturing, object detection can
help automate inventory management and quality control processes. Similarly, in
healthcare, object detection is used in medical imaging to identify tumors, organs,
and other anomalies.

5
 TABLE OF CONTENTS

CHAPTER NO. TITLE PAG

E
NO.
SYNOPSIS
LIST OF TABLES LIST iii
OF FIGURES v
i
i
v
i
i
i

1. INTRODUCTION 1
1.1. PROBLEM STATEMENT 2
1.2. OBJECTIVE 2
1.2.1.SCOPE 3
1.3. EXISTING SOFTWARE 4
1.4. MOTIVATION 4

2. BACKGROUND AND RELATED WORK 6

3. HARDWARE AND SOFTWARE REQUIREMENTS 10

3.1 HARDWARE REQUIREMENT 10

3.2 SOFTWARE REQUIREMENT 10

3.3 TESTING TECHNOLOGY 10

4. SDLC METHODOLOGIES 11
1. AVAILABLE SDLC MODELS 11
1.1. WATERFALL 11
1.2. RAD MODEL 12
1.3. SPIRAL MODEL 12
1.4. INCREMENTAL MODEL 13
2. MODEL USED IN PROJECT: AGILE 14

5. APPLICATION ARCHITECTURE 16

1. MVVM 6 17
 2. PHASES OF PROJECT 18
3. USE CASE DIAGRAM 18
5.4 CLASS DIAGRAM 19

6. REFERENCES 20

7
 LIST OF TABLES

CHAPTER NO. TABLE NO. TITLE PAGE NO.

2 TABLE 2.1 COMPARISON OF VARIOUS 6

METHODOLOGY

7
 LIST OF FIGURES

CHAPTER NO. TITLE PAGE NO.

4 Figure 4.1. Waterfall Model 11

4 Figure 4.2. RAD Model 12
4 Figure 4.3. Spiral Model 13
4 Figure 4.4. Incremental Model 14
4 Figure 4.5. Agile Model 15
5 Figure 5.1. Application Architecture 16
5 Figure 5.2. MVVM Model 17
5 Figure 5.3. Use Case Diagram 18
5 Figure 5.4. Class Diagram 19