You are on page 1of 13

Mini Project Report on

FACE DETECTION USING DEEP LEARNING

Submitted in partial fulfilment of the requirement for the award of the degree of

BACHELOR OF TECHNOLOGY

IN

COMPUTER SCIENCE & ENGINEERING

Submitted by:

AYUSHI BHARDWAJ 2016700

Under the Mentorship of

Mr. ANKIT GUPTA


Assistant Proffessor

Department of Computer Science and Engineering


Graphic Era (Deemed to be University)
Dehradun, Uttarakhand
July-2023
CANDIDATE’S DECLARATION

I hereby certify that the work which is being presented in the project report entitled “Facial
Detection using Deep Learning” in partial fulfilment of the requirements for the award of
the Degree of Bachelor of Technology in Computer Science and Engineering of the Graphic
Era (Deemed to be University), Dehradun shall be carried out by the under the mentorship of
Mr. Ankit Gupta, Assistant Proffessor, Department of Computer Science and Engineering,
Graphic Era (Deemed to be University), Dehradun.

Ayushi Bhardwaj 2016700


Table of Contents

Chapter Description Page No.


No.
Chapter 1 Introduction
Chapter 2 Literature Survey
Chapter 3 Methodology
Chapter 4 Result and Discussion
Chapter 5 Conclusion and Future Work
References
Chapter 1

Introduction

1. Introduction

In this time of vast expansion of internet usage ,face detection has been proved to be one of
the most crucial and fastest form of biometric verification. one of the best methods to
implement this is to use Machine learning techniques along with image processing methods .
Machine learning have been very helpful for implementing such crucial systems. This project
uses Deep learning Algorithms to compare a captured image(live) and the images stored in
database to verify the identity. This type of face recognition system are also crucial from
security perspective. This type of systems may help in face tracking, criminal detection,
fingerprint matching etc.Uses such AI techniques one can even detect emotions of an
individual.

Face detection system uses AI aka Artificial Intelligence techniques to identify any object or
a person from video or an image . This type of system can be built using a combination of AI
techniques like deep learning, Image processing and computer vision. This type of face
detection systems have been widely popular in a wide variety of applications like smart
phone unlocking, authentication at airports, door unlocking, security related and so on.
Figure 1.1 Basic outlay about how facial detection system works

Using deep learning, a computer model learns to perform classification tasks using the vast
database provide to that model and comparing the real time video, images to the already
stored database. Their accuracy may vary depending upon the amount of data provided to the
model. Sometimes this accuracy may exceed the human level performance. Therefore the
training of the data plays a major role in the accuracy of the system.

Deep Learning helps computer to perform basic daily life tasks like voice control in devices
like mobile phone, television, speakers .it is also the key technology behind the latest
driverless cars technology. Therefore due to such key features it has been used widely to
develop system that will help the human to perform such basic tasks.

How does Deep Learning works?

The term “deep” refers to number of layers in the neural networks. Most Deep Learning
methods use neural network architectures, therefore often referred as Deep Neural Networks.
This model is trained using a vast amount of labelled data and neural network architecture
that only uses data and does not require any manual efforts.
Deep Learning is a specialized form of Machine Learning. It performs end to end learning
where we provide network large amount of data and a specific task to perform such as
comparison or classification. The key feature of Deep learning is that its accuracy is directly
proportional to the amount of data provided as database, more the data high will be its

accuracy.

Figure 1.2 Fully Connected Artificial Neural Network

Chapter 2

Literature Survey
This section gives an overview to the major recognition techniques that apply mostly to faces
, along with their advantages and disadvantages . The methods are eigenfaces (eigenfeatures),
neural networks, dynamic link architecture, hidden Markov model geometrical feature
matching, and template matching.

1.Eigenface

Eigenface is one of the most thoroughly investigated approaches to face recognition. It is


also known as eigen picture .It used principal component analysis to efficiently represent
pictures of faces. They argued that any face images can be easily reconstructed using a small
collection of weights for each face and a standard face (eigen picture). The weights were
obtained by projecting the face image onto the eigen picture.

2.Neural Networks
As the neural networks supports non linearity in the network, therefore the feature
extraction step in case of neural networks will be far more efficient. One of the first
artificial neural networks techniques used for face recognition is a single layer adaptive
network was called WISARD .For face detection and recognition multiple layer CNN have to
be applied .

3.Graph Matching

Graph matching is another approach towards face recognition. This approach used a
dynamically linked structured which further uses elastic graph matching to find the closest
matching one. This type of Dynamically linked architecture is basically an extension to the
classical ANN(artificial neural networks).However this kind of matching process is
computationally rather more expensive as compared to the earlier approaches.

4.Geometrical Feature Matching


Geometrical feature matching technique is based on the computing of a set of features of a
face. The fact that face recognition is possible even at a low resolution like 8x6 pixels when
facial features are hardly visible in detail, implies that this technique is highly accurate for
recognition. The overall configuration can be described by a vector representing the position
and size of the main facial features, such as nose, mouth, eyes and shape of the face.

5.Template Matching

In template Matching Technique test image is represented as a 2D array and then it is


compared using a suitable metric . Their can be various other more sophisticated versions
of using template matching for face recognition for example using different viewpoints to
represent an individual’s face .
Chapter 3

Methodology
To build a face detection statement using deep learning , we need to work on three distinct

phases.

 Data gathering and face detection

 Training the recognizer

 The final step, Face Recognition


Figure 3.1 various phases of face detection system

Face Detection system uses Deep learning techniques as well as image processing to provide
the required result , that is , to find out a face or to distinguish it from the landscape or
building or other human body parts . The foremost step before beginning the detection is to
improvise the image quality and remove all those interfernces that might disturb or be an
obstacle in detection.

Once the captured image is processed for better quality and better understanding , the next
step begins by detecting the various important feature of face which might help in in face
recognition . so this algorithm typically starts with eye searching as it is comparatively
easier. This may further continue by searching for mouth, nose ,iris etc. once the facial
features are detected ,the face has been detected.

Once the real time face from video or image is captured , next task is to perform recognition .
For this the system needs to be provided wit high amount of data, as accuracy of the system is
directly proportional to the amount of trained data provided to the model. The more the data
is provided , higher will be its accuracy of face recognition. Therefore the algorithm is trained
on large labelled data sets, this training improves the models ability to recognize face.
Once the algorithm is trained with database and the face is detected from the real time
image/video the next step is comparison or classification . On the basis of best matching
results , the model will decide that to whom that picture matches and display the labelled
name from database to the real time video.one can also display the accuracy percentage, that
is, percentage of how much the real time image matches to the image stored in database of
the resulting identity.

Figure 3.2 working of face recognition system

Chapter 4

Result and Discussion

Here are few of my output screenshots , where I have not only detected the face but also tried
to recognize them as well.
Chapter 5

Conclusion and Future Work

Hereby I, conclude this project of face detection using deep learning method. The face can be
clearly detected from the screen no matter the number of faces or number of obstacles, it will
easily and accurately detect the face .
However I have also implemented face recognition in this project but as the main focus was
to detect the face , so I have not provided much database to my model. So this type of face
detection model can be further used as a basis model for face recognition model as well. And
the accuracy of face recognition can be easily increased by increasing the trained and labelled
data to the model.

References

[1] https://www.geeksforgeeks.org/introduction-deep-learning/

[2] https://web.stanford.edu/class/ee368/Project_03/Project/reports/ee368group02.pdf

[3] https://www.researchgate.net/publication/233864740_Face_Recognition_A_Literature_Review

[4] https://towardsdatascience.com/real-time-face-recognition-an-end-to-end-project-b738bb0f7348

[5] https://rcciit.org/students_projects/projects/cse/2018/GR2.pdf
[6]
https://www.academia.edu/38912204/A_PROJECT_REPORT_ON_FACE_RECOGNITION_SYSTE
M_WITH_FACE_DETECTION

You might also like