SlideShare a Scribd company logo
THATCHER EFFECT AND
FACE RECOGNITION
By: Erfan Alimohammadi
Supervisor: Prof. Ebrahimi Moghaddam
Date: Summer 2020
MOTIVATION AND BACKGROUND
WHAT IS THATCHER EFFECT?
DATASET
THE FIRST PROBLEM
THE SECOND PROBLEM
CONCLUSION
REFERENCES
TABLE OF CONTENTS
01
02
03
04
05
06
07
MOTIVATION AND BACKGROUND
01
What does motivate us to solve
these kind of problems?
MOTIVATION
AND BACKGROUND
How do computers see optical illusions?
Do they see them in the way we see them?
These horizontal
lines are parallel.
There is only one color
in the vertical rectangle.
WHAT IS THATCHER EFFECT?
It’s an optical illusion!
02
WHAT IS
THATCHER EFFECT?
You’re given a face image.
1. Flip the left eye vertically.
2. Flip the right eye vertically.
3. Flip the mouth vertically.
4. Flip whole image vertically.
Now, you have a “thatcherized” image
which causes an optical illusion.
You’ll notice it when you turn the
image upside-down.
Turn your phone
upside-down
A thatcherized image The same image, but
upside-down
DATASET
What are some popular face
datasets and which one should we
use for solving our problems?
03
DATASET
● LFW
● CASIA WebFace
● CelebA
● MS-CELEB-1M
● VGGFace2 (2GB): 500 identities
● VGGFace2 (40GB): 9000 identities
After selecting VGGFace2 among these well-known datasets, I decided to
crop its images using MTCNN face detection algorithm.
THATCHERIZED DATASET
Detecting position of the eyes and the mouth, has helped me to
automatically thatcherize each image of my original dataset separately.
dlib library 68 facial
landmarks detection
THE FIRST PROBLEM
04
Is this image thatcherized, or not?
There are two normal faces and two thatcherized faces here.
You’ll only notice it when you flip all four images vertically.
WHICH ONES ARE THATCHERIZED?
WHICH ONES ARE NORMAL?
APPROACH TO
THE FIRST PROBLEM
I have used a neural network for solving this
binary classification problem:
1. 5 Convolutional layers with ReLU
2. Maxpooling layer
3. Dropout
4. 4 Fully-connected layers with ReLU
5. Apply sigmoid on the output
And, the used loss function was binary cross
entropy.
training accuracy 0.994
test accuracy 0.999
RESULT ON VGGFACE2 DATASET (2GB)
training accuracy 0.999
test accuracy 0.999
RESULT ON VGGFACE2 DATASET (40GB)
THE SECOND PROBLEM
Effects of Thatcher illusion on
face recognition
05
APPROACH TO
THE SECOND PROBLEM
Inception ResNet v1 architecture
has been used here with cross
entropy loss function.
Inception ResNet v1 structure
training set thatcherized
images share
0% 10% 20%
training accuracy 0.910 0.955 0.978
test accuracy 0.820 0.945 0.968
RESULT ON VGGFACE2 (2GB)
training set thatcherized
images share
0%
training accuracy 0.890
test accuracy 0.735
RESULT ON VGGFACE2 DATASET (40GB)
CONCLUSION
What’s next?
06
CONCLUSION AND FUTURE WORKS
● Training on thatcherized images increases our network robustness in
face recognition problems.
● Generative networks can be used for creating thatcherized images.
● We may use Grad-CAM for more visualizations.
● We can test other optical illusions too.
● Future researches can be continued in the field of cognitive sciences.
REFERENCES
07
Where do images, datasets,
libraries, etc. come from?
REFERENCES
● Slide images from Wikipedia and CelebA dataset
● Q. Cao, L. Shen, W. Xie, O. M. Parkhi, A. Zisserman. VGGFace2: A
dataset for recognising face across pose and age, International
Conference on Automatic Face and Gesture Recognition, 2018
● K. Zhang, Z. Zhang, Z. Li and Y. Qiao. Joint Face Detection and
Alignment Using Multitask Cascaded Convolutional Networks,
IEEE Signal Processing Letters, 2016
● V. Kazemi, J. Sullivan: One millisecond face alignment with an
ensemble of regression trees, 2014
● Tim Esler’s FaceNet PyTorch repository:
https://github.com/timesler/facenet-pytorch
THE END!
With special thanks to:
Prof. Mohsen Ebrahimi Moghaddam
Mr. Ali Nazari

More Related Content

Similar to Bachelor's Project.pdf

F04613040
F04613040F04613040
F04613040
IOSR-JEN
 
Infarec
InfarecInfarec
Infarec
sparwaiz
 
jefferson-mae Masked Autoencoders based Pretraining
jefferson-mae Masked Autoencoders based Pretrainingjefferson-mae Masked Autoencoders based Pretraining
jefferson-mae Masked Autoencoders based Pretraining
cevesom156
 
COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...
COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...
COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...
SaiPrakash106
 
Face recognition across non uniform motion
Face recognition across non uniform motionFace recognition across non uniform motion
Face recognition across non uniform motion
jpstudcorner
 
IRJET- IoT based Door Lock and Unlock System using Face Recognition
IRJET- IoT based Door Lock and Unlock System using Face RecognitionIRJET- IoT based Door Lock and Unlock System using Face Recognition
IRJET- IoT based Door Lock and Unlock System using Face Recognition
IRJET Journal
 
IRJET- ATM Security using Machine Learning
IRJET- ATM Security using Machine LearningIRJET- ATM Security using Machine Learning
IRJET- ATM Security using Machine Learning
IRJET Journal
 
Introducing Set Of Internal Parameters For Laplacian Faces
Introducing Set Of Internal Parameters For Laplacian FacesIntroducing Set Of Internal Parameters For Laplacian Faces
Introducing Set Of Internal Parameters For Laplacian Faces
Vision and Pattern Recognition Systems
 
vision correcting display
vision correcting displayvision correcting display
vision correcting display
Disha Tiwari
 
Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์
Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์
Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์
BAINIDA
 
PBL presentation p2.pptx
PBL presentation p2.pptxPBL presentation p2.pptx
PBL presentation p2.pptx
Tony383416
 
Report face recognition : ArganRecogn
Report face recognition :  ArganRecognReport face recognition :  ArganRecogn
Report face recognition : ArganRecogn
Ilyas CHAOUA
 
Digital Future of the Surgery: Brining the Innovation of Digital Technology i...
Digital Future of the Surgery: Brining the Innovation of Digital Technology i...Digital Future of the Surgery: Brining the Innovation of Digital Technology i...
Digital Future of the Surgery: Brining the Innovation of Digital Technology i...
Yoon Sup Choi
 
Face and Eye Detection Varying Scenarios With Haar Classifier_2015
Face and Eye Detection Varying Scenarios With Haar Classifier_2015Face and Eye Detection Varying Scenarios With Haar Classifier_2015
Face and Eye Detection Varying Scenarios With Haar Classifier_2015
Showrav Mazumder
 
Laureate Online Education Internet and Multimedia Technolog.docx
Laureate Online Education    Internet and Multimedia Technolog.docxLaureate Online Education    Internet and Multimedia Technolog.docx
Laureate Online Education Internet and Multimedia Technolog.docx
DIPESH30
 
Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...
Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...
Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...
sugiuralab
 
Face detection and tracking in a video sequence
Face detection and tracking in a video sequenceFace detection and tracking in a video sequence
Face detection and tracking in a video sequence
Karthik G N
 
Face Detection And Tracking
Face Detection And TrackingFace Detection And Tracking
Face Detection And Tracking
NarayanlalMenariya
 
A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...
A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...
A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...
mlaij
 
Machine Learning Model for Gender Detection
Machine Learning Model for Gender DetectionMachine Learning Model for Gender Detection
Machine Learning Model for Gender Detection
TecnoIncentive
 

Similar to Bachelor's Project.pdf (20)

F04613040
F04613040F04613040
F04613040
 
Infarec
InfarecInfarec
Infarec
 
jefferson-mae Masked Autoencoders based Pretraining
jefferson-mae Masked Autoencoders based Pretrainingjefferson-mae Masked Autoencoders based Pretraining
jefferson-mae Masked Autoencoders based Pretraining
 
COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...
COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...
COVID-19-Preventions-Control-System and Unconstrained Face-mask and Face-hand...
 
Face recognition across non uniform motion
Face recognition across non uniform motionFace recognition across non uniform motion
Face recognition across non uniform motion
 
IRJET- IoT based Door Lock and Unlock System using Face Recognition
IRJET- IoT based Door Lock and Unlock System using Face RecognitionIRJET- IoT based Door Lock and Unlock System using Face Recognition
IRJET- IoT based Door Lock and Unlock System using Face Recognition
 
IRJET- ATM Security using Machine Learning
IRJET- ATM Security using Machine LearningIRJET- ATM Security using Machine Learning
IRJET- ATM Security using Machine Learning
 
Introducing Set Of Internal Parameters For Laplacian Faces
Introducing Set Of Internal Parameters For Laplacian FacesIntroducing Set Of Internal Parameters For Laplacian Faces
Introducing Set Of Internal Parameters For Laplacian Faces
 
vision correcting display
vision correcting displayvision correcting display
vision correcting display
 
Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์
Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์
Face recognition and modeling โดย ผศ.ดร.ธนาสัย สุคนธ์พันธุ์
 
PBL presentation p2.pptx
PBL presentation p2.pptxPBL presentation p2.pptx
PBL presentation p2.pptx
 
Report face recognition : ArganRecogn
Report face recognition :  ArganRecognReport face recognition :  ArganRecogn
Report face recognition : ArganRecogn
 
Digital Future of the Surgery: Brining the Innovation of Digital Technology i...
Digital Future of the Surgery: Brining the Innovation of Digital Technology i...Digital Future of the Surgery: Brining the Innovation of Digital Technology i...
Digital Future of the Surgery: Brining the Innovation of Digital Technology i...
 
Face and Eye Detection Varying Scenarios With Haar Classifier_2015
Face and Eye Detection Varying Scenarios With Haar Classifier_2015Face and Eye Detection Varying Scenarios With Haar Classifier_2015
Face and Eye Detection Varying Scenarios With Haar Classifier_2015
 
Laureate Online Education Internet and Multimedia Technolog.docx
Laureate Online Education    Internet and Multimedia Technolog.docxLaureate Online Education    Internet and Multimedia Technolog.docx
Laureate Online Education Internet and Multimedia Technolog.docx
 
Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...
Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...
Automatic Eyeglasses Replacement for a 3D Virtual Try-on System (AH2019 Short...
 
Face detection and tracking in a video sequence
Face detection and tracking in a video sequenceFace detection and tracking in a video sequence
Face detection and tracking in a video sequence
 
Face Detection And Tracking
Face Detection And TrackingFace Detection And Tracking
Face Detection And Tracking
 
A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...
A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...
A COMPREHENSIVE STUDY ON OCCLUSION INVARIANT FACE RECOGNITION UNDER FACE MASK...
 
Machine Learning Model for Gender Detection
Machine Learning Model for Gender DetectionMachine Learning Model for Gender Detection
Machine Learning Model for Gender Detection
 

Recently uploaded

The Intersection between Competition and Data Privacy – COLANGELO – June 2024...
The Intersection between Competition and Data Privacy – COLANGELO – June 2024...The Intersection between Competition and Data Privacy – COLANGELO – June 2024...
The Intersection between Competition and Data Privacy – COLANGELO – June 2024...
OECD Directorate for Financial and Enterprise Affairs
 
怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样
怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样
怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样
kekzed
 
BRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdf
BRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdfBRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdf
BRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdf
Robin Haunschild
 
Competition and Regulation in Professions and Occupations – OECD – June 2024 ...
Competition and Regulation in Professions and Occupations – OECD – June 2024 ...Competition and Regulation in Professions and Occupations – OECD – June 2024 ...
Competition and Regulation in Professions and Occupations – OECD – June 2024 ...
OECD Directorate for Financial and Enterprise Affairs
 
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
OECD Directorate for Financial and Enterprise Affairs
 
Why Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdf
Why Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdfWhy Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdf
Why Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdf
Ben Linders
 
Artificial Intelligence, Data and Competition – OECD – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – OECD – June 2024 OECD discussionArtificial Intelligence, Data and Competition – OECD – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – OECD – June 2024 OECD discussion
OECD Directorate for Financial and Enterprise Affairs
 
Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...
Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...
Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...
Suzanne Lagerweij
 
Pro-competitive Industrial Policy – OECD – June 2024 OECD discussion
Pro-competitive Industrial Policy – OECD – June 2024 OECD discussionPro-competitive Industrial Policy – OECD – June 2024 OECD discussion
Pro-competitive Industrial Policy – OECD – June 2024 OECD discussion
OECD Directorate for Financial and Enterprise Affairs
 
The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...
The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...
The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...
OECD Directorate for Financial and Enterprise Affairs
 
原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样
原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样
原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样
gpww3sf4
 
Disaster Management project for holidays homework and other uses
Disaster Management project for holidays homework and other usesDisaster Management project for holidays homework and other uses
Disaster Management project for holidays homework and other uses
RIDHIMAGARG21
 
Artificial Intelligence, Data and Competition – LIM – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – LIM – June 2024 OECD discussionArtificial Intelligence, Data and Competition – LIM – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – LIM – June 2024 OECD discussion
OECD Directorate for Financial and Enterprise Affairs
 
Using-Presentation-Software-to-the-Fullf.pptx
Using-Presentation-Software-to-the-Fullf.pptxUsing-Presentation-Software-to-the-Fullf.pptx
Using-Presentation-Software-to-the-Fullf.pptx
kainatfatyma9
 
The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...
The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...
The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...
OECD Directorate for Financial and Enterprise Affairs
 
IEEE CIS Webinar Sustainable futures.pdf
IEEE CIS Webinar Sustainable futures.pdfIEEE CIS Webinar Sustainable futures.pdf
IEEE CIS Webinar Sustainable futures.pdf
Claudio Gallicchio
 
Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...
Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...
Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...
OECD Directorate for Financial and Enterprise Affairs
 
The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...
The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...
The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...
OECD Directorate for Financial and Enterprise Affairs
 
Competition and Regulation in Professions and Occupations – ROBSON – June 202...
Competition and Regulation in Professions and Occupations – ROBSON – June 202...Competition and Regulation in Professions and Occupations – ROBSON – June 202...
Competition and Regulation in Professions and Occupations – ROBSON – June 202...
OECD Directorate for Financial and Enterprise Affairs
 
Pro-competitive Industrial Policy – LANE – June 2024 OECD discussion
Pro-competitive Industrial Policy – LANE – June 2024 OECD discussionPro-competitive Industrial Policy – LANE – June 2024 OECD discussion
Pro-competitive Industrial Policy – LANE – June 2024 OECD discussion
OECD Directorate for Financial and Enterprise Affairs
 

Recently uploaded (20)

The Intersection between Competition and Data Privacy – COLANGELO – June 2024...
The Intersection between Competition and Data Privacy – COLANGELO – June 2024...The Intersection between Competition and Data Privacy – COLANGELO – June 2024...
The Intersection between Competition and Data Privacy – COLANGELO – June 2024...
 
怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样
怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样
怎么办理(lincoln学位证书)英国林肯大学毕业证文凭学位证书原版一模一样
 
BRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdf
BRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdfBRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdf
BRIC_2024_2024-06-06-11:30-haunschild_archival_version.pdf
 
Competition and Regulation in Professions and Occupations – OECD – June 2024 ...
Competition and Regulation in Professions and Occupations – OECD – June 2024 ...Competition and Regulation in Professions and Occupations – OECD – June 2024 ...
Competition and Regulation in Professions and Occupations – OECD – June 2024 ...
 
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
Artificial Intelligence, Data and Competition – SCHREPEL – June 2024 OECD dis...
 
Why Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdf
Why Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdfWhy Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdf
Why Psychological Safety Matters for Software Teams - ACE 2024 - Ben Linders.pdf
 
Artificial Intelligence, Data and Competition – OECD – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – OECD – June 2024 OECD discussionArtificial Intelligence, Data and Competition – OECD – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – OECD – June 2024 OECD discussion
 
Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...
Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...
Suzanne Lagerweij - Influence Without Power - Why Empathy is Your Best Friend...
 
Pro-competitive Industrial Policy – OECD – June 2024 OECD discussion
Pro-competitive Industrial Policy – OECD – June 2024 OECD discussionPro-competitive Industrial Policy – OECD – June 2024 OECD discussion
Pro-competitive Industrial Policy – OECD – June 2024 OECD discussion
 
The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...
The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...
The Intersection between Competition and Data Privacy – OECD – June 2024 OECD...
 
原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样
原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样
原版制作贝德福特大学毕业证(bedfordhire毕业证)硕士文凭原版一模一样
 
Disaster Management project for holidays homework and other uses
Disaster Management project for holidays homework and other usesDisaster Management project for holidays homework and other uses
Disaster Management project for holidays homework and other uses
 
Artificial Intelligence, Data and Competition – LIM – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – LIM – June 2024 OECD discussionArtificial Intelligence, Data and Competition – LIM – June 2024 OECD discussion
Artificial Intelligence, Data and Competition – LIM – June 2024 OECD discussion
 
Using-Presentation-Software-to-the-Fullf.pptx
Using-Presentation-Software-to-the-Fullf.pptxUsing-Presentation-Software-to-the-Fullf.pptx
Using-Presentation-Software-to-the-Fullf.pptx
 
The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...
The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...
The Intersection between Competition and Data Privacy – KEMP – June 2024 OECD...
 
IEEE CIS Webinar Sustainable futures.pdf
IEEE CIS Webinar Sustainable futures.pdfIEEE CIS Webinar Sustainable futures.pdf
IEEE CIS Webinar Sustainable futures.pdf
 
Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...
Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...
Artificial Intelligence, Data and Competition – ČORBA – June 2024 OECD discus...
 
The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...
The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...
The Intersection between Competition and Data Privacy – CAPEL – June 2024 OEC...
 
Competition and Regulation in Professions and Occupations – ROBSON – June 202...
Competition and Regulation in Professions and Occupations – ROBSON – June 202...Competition and Regulation in Professions and Occupations – ROBSON – June 202...
Competition and Regulation in Professions and Occupations – ROBSON – June 202...
 
Pro-competitive Industrial Policy – LANE – June 2024 OECD discussion
Pro-competitive Industrial Policy – LANE – June 2024 OECD discussionPro-competitive Industrial Policy – LANE – June 2024 OECD discussion
Pro-competitive Industrial Policy – LANE – June 2024 OECD discussion
 

Bachelor's Project.pdf

  • 1. THATCHER EFFECT AND FACE RECOGNITION By: Erfan Alimohammadi Supervisor: Prof. Ebrahimi Moghaddam Date: Summer 2020
  • 2. MOTIVATION AND BACKGROUND WHAT IS THATCHER EFFECT? DATASET THE FIRST PROBLEM THE SECOND PROBLEM CONCLUSION REFERENCES TABLE OF CONTENTS 01 02 03 04 05 06 07
  • 3. MOTIVATION AND BACKGROUND 01 What does motivate us to solve these kind of problems?
  • 4. MOTIVATION AND BACKGROUND How do computers see optical illusions? Do they see them in the way we see them? These horizontal lines are parallel. There is only one color in the vertical rectangle.
  • 5. WHAT IS THATCHER EFFECT? It’s an optical illusion! 02
  • 6. WHAT IS THATCHER EFFECT? You’re given a face image. 1. Flip the left eye vertically. 2. Flip the right eye vertically. 3. Flip the mouth vertically. 4. Flip whole image vertically. Now, you have a “thatcherized” image which causes an optical illusion. You’ll notice it when you turn the image upside-down. Turn your phone upside-down A thatcherized image The same image, but upside-down
  • 7. DATASET What are some popular face datasets and which one should we use for solving our problems? 03
  • 8. DATASET ● LFW ● CASIA WebFace ● CelebA ● MS-CELEB-1M ● VGGFace2 (2GB): 500 identities ● VGGFace2 (40GB): 9000 identities After selecting VGGFace2 among these well-known datasets, I decided to crop its images using MTCNN face detection algorithm.
  • 9. THATCHERIZED DATASET Detecting position of the eyes and the mouth, has helped me to automatically thatcherize each image of my original dataset separately. dlib library 68 facial landmarks detection
  • 10. THE FIRST PROBLEM 04 Is this image thatcherized, or not?
  • 11. There are two normal faces and two thatcherized faces here. You’ll only notice it when you flip all four images vertically. WHICH ONES ARE THATCHERIZED? WHICH ONES ARE NORMAL?
  • 12. APPROACH TO THE FIRST PROBLEM I have used a neural network for solving this binary classification problem: 1. 5 Convolutional layers with ReLU 2. Maxpooling layer 3. Dropout 4. 4 Fully-connected layers with ReLU 5. Apply sigmoid on the output And, the used loss function was binary cross entropy.
  • 13. training accuracy 0.994 test accuracy 0.999 RESULT ON VGGFACE2 DATASET (2GB)
  • 14. training accuracy 0.999 test accuracy 0.999 RESULT ON VGGFACE2 DATASET (40GB)
  • 15. THE SECOND PROBLEM Effects of Thatcher illusion on face recognition 05
  • 16. APPROACH TO THE SECOND PROBLEM Inception ResNet v1 architecture has been used here with cross entropy loss function. Inception ResNet v1 structure
  • 17. training set thatcherized images share 0% 10% 20% training accuracy 0.910 0.955 0.978 test accuracy 0.820 0.945 0.968 RESULT ON VGGFACE2 (2GB)
  • 18. training set thatcherized images share 0% training accuracy 0.890 test accuracy 0.735 RESULT ON VGGFACE2 DATASET (40GB)
  • 20. CONCLUSION AND FUTURE WORKS ● Training on thatcherized images increases our network robustness in face recognition problems. ● Generative networks can be used for creating thatcherized images. ● We may use Grad-CAM for more visualizations. ● We can test other optical illusions too. ● Future researches can be continued in the field of cognitive sciences.
  • 21. REFERENCES 07 Where do images, datasets, libraries, etc. come from?
  • 22. REFERENCES ● Slide images from Wikipedia and CelebA dataset ● Q. Cao, L. Shen, W. Xie, O. M. Parkhi, A. Zisserman. VGGFace2: A dataset for recognising face across pose and age, International Conference on Automatic Face and Gesture Recognition, 2018 ● K. Zhang, Z. Zhang, Z. Li and Y. Qiao. Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks, IEEE Signal Processing Letters, 2016 ● V. Kazemi, J. Sullivan: One millisecond face alignment with an ensemble of regression trees, 2014 ● Tim Esler’s FaceNet PyTorch repository: https://github.com/timesler/facenet-pytorch
  • 23. THE END! With special thanks to: Prof. Mohsen Ebrahimi Moghaddam Mr. Ali Nazari