SlideShare a Scribd company logo
1 of 61
B.Sc. Final Project – Mohammad Pooya Malek 2
University of IRIB
Faculty of Broadcast Engineering
A Thesis Presented for the Degree of
Bachelor of Science in Telecommunications Engineering
Image Inpainting
Based on Deep Learning
By
Mohammad Pooya Malek
Supervisor Master Reviewer
Dr. Azam Bastanfard Dr. Rahil Mahdian
Summer 2018
B.Sc. Final Project – Mohammad Pooya Malek 2
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Aim of Thesis:
 Censorships
Occluded Images
Appropriate reconstruction
To Complete the image which has been destroyed
To repair old Images
B.Sc. Final Project – Mohammad Pooya Malek
Index
3
Introduction
Related work
Problems
Recommended Algorithm (Proposed Model)
Samples From Model
Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
Introduction
4
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
Introduction
5
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Computer Vision (Machine Vision)
• Artificial Intelligence (AI)
• Deep Learning
• Machine Learning
B.Sc. Final Project – Mohammad Pooya Malek
6
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
7
Neural
Network
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
8
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Neural Networks: Architectures
“Fully-connected” layers
“2-layer Neural Net”, or
“1-hidden-layer Neural Net”
“3-layer Neural Net”, or
“2-hidden-layer Neural Net”
B.Sc. Final Project – Mohammad Pooya Malek
9
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Neural Networks: Architectures
B.Sc. Final Project – Mohammad Pooya Malek
10
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
32
32
3
Convolution Layer
5x5x3 filter
32x32x3 image
B.Sc. Final Project – Mohammad Pooya Malek
11
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
32
32
3
Convolution Layer
32x32x3 image
5x5x3 filter
convolve (slide) over all spatial
locations
Activation map
1
28
28
B.Sc. Final Project – Mohammad Pooya Malek
12
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
32
32
3
Convolution Layer
32x32x3 image
5x5x3 filter
convolve (slide) over all spatial
locations
activation maps
2
28
28
consider a second, green filter
1
B.Sc. Final Project – Mohammad Pooya Malek
13
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
32
32
3
Convolution Layer
activation maps
6
28
28
For example, if we had 6, 5x5 filters, we’ll get 6 separate activation maps:
We stack these up to get a “new image” of size 28x28x6!
B.Sc. Final Project – Mohammad Pooya Malek
14
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Preview: ConvNet is a sequence of Convolutional Layers, interspersed with activation
functions
32
32
3
CONV,
ReLU
e.g. 6
5x5x3
filters 28
28
6
CONV,
ReLU
e.g. 10
5x5x6
filters
CONV,
ReLU
….
10
24
24
B.Sc. Final Project – Mohammad Pooya Malek
15
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
16
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Convolutional Layer
B.Sc. Final Project – Mohammad Pooya Malek
17
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Convolutional Layer
http://setosa.io/ev/image-kernels/
B.Sc. Final Project – Mohammad Pooya Malek
18
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Activation Functions
Sigmoid
tanh tanh(x)
ReLU max(0,x)
Leaky ReLU
max(0.1x, x)
ELU
B.Sc. Final Project – Mohammad Pooya Malek
19
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Activation Functions
ReLU
(Rectified Linear Unit)
- Computes f(x) = max(0,x)
- Does not saturate (in +region)
- Very computationally efficient
- Converges much faster than sigmoid/tanh
in practice (e.g. 6x)
- Not zero-centered output
- And the gradient when x < 0 is 0
B.Sc. Final Project – Mohammad Pooya Malek
20
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Activation Functions
Leaky ReLU
- Does not saturate
- Computationally efficient
- Converges much faster than
sigmoid/tanh in practice! (e.g. 6x)
- will not “die”.
Parametric Rectifier (PReLU)
backprop into alpha
(parameter)
B.Sc. Final Project – Mohammad Pooya Malek
21
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Activation Functions
Exponential Linear Units (ELU)
- All benefits of ReLU
- Does not die
- Closer to zero mean outputs
- Computation requires exp()
B.Sc. Final Project – Mohammad Pooya Malek 22
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
ReLU (Rectified Linear Units)
max(0,x)
B.Sc. Final Project – Mohammad Pooya Malek
23
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Pooling layer
B.Sc. Final Project – Mohammad Pooya Malek
24
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Fully
Connected
layer
B.Sc. Final Project – Mohammad Pooya Malek
25
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
LeNet-5, a pioneering convolutional network by LeCun. in 1998
classifies digits
Recognize hand-written numbers on 32x32 pixel images.
LeNet Architecture
B.Sc. Final Project – Mohammad Pooya Malek
26
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
AlexNet Architecture
Alex Krizhevsky, Geoffrey Hinton, and Ilya Sutskever
ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2012 Winner
AlexNet uses ReLU(Rectified Linear Unit) for the non-linear part,
instead of a Tanh or Sigmoid function
The advantage of Using ReLU over sigmoid is that it trains much faster
Full (simplified) AlexNet architecture:
[227x227x3] INPUT
[55x55x96] CONV1: 96 11x11 filters at stride 4, pad 0
[27x27x96] MAX POOL1: 3x3 filters at stride 2
[27x27x96] NORM1: Normalization layer
[27x27x256] CONV2: 256 5x5 filters at stride 1, pad 2
[13x13x256] MAX POOL2: 3x3 filters at stride 2
[13x13x256] NORM2: Normalization layer
[13x13x384] CONV3: 384 3x3 filters at stride 1, pad 1
[13x13x384] CONV4: 384 3x3 filters at stride 1, pad 1
[13x13x256] CONV5: 256 3x3 filters at stride 1, pad 1
[6x6x256] MAX POOL3: 3x3 filters at stride 2
[4096] FC6: 4096 neurons
[4096] FC7: 4096 neurons
[1000] FC8: 1000 neurons (class scores)
B.Sc. Final Project – Mohammad Pooya Malek
27
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
VGGNet16 Architecture
Visual Geometry Group
Made improvement over AlexNet
Increases the depth of the network which enable it to
learn more complex features
B.Sc. Final Project – Mohammad Pooya Malek
28
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
VGGnet19 Architecture
B.Sc. Final Project – Mohammad Pooya Malek
29
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
CNN Architectures
B.Sc. Final Project – Mohammad Pooya Malek 30
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Related work
Criminisi Algorithm
B.Sc. Final Project – Mohammad Pooya Malek
31
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
32
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Problems
B.Sc. Final Project – Mohammad Pooya Malek
33
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Algorithm
Supervised vs
Unsupervised
B.Sc. Final Project – Mohammad Pooya Malek
34
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
OpenCV (Open Source Computer Vision Library)
BSD license: free for both academic and commercial usage
Open Source: C++, Python and Java interfaces
Cross-platform: Windows, Linux, Mac OS, iOS and Android.
focus on real-time applications
Originally developed by Intel 1999; supported by Willow Garage
maintained by Itseez; Adopted all around the world: 47 thousand user and community
downloads exceeding 14 million times; last release: 3.4.0 - 23 December 2017
Tensorflow
BSD license: free for both academic and commercial usage
Open Source: Python, C++, CUDA
Cross-platform: Linux, macOS, Windows, Android, iOS and website.
open source software library for high performance numerical computation
TensorFlow was developed by the Google Brain team in 2011
Initial released on November 9, 2015; 2 years ago
TensorFlow can run on multiple CPUs and GPUs
last release: August 8, 2018; 34 days ago
B.Sc. Final Project – Mohammad Pooya Malek
35
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Supervised vs Unsupervised
Supervised Learning
Data: (x, y)
x is data, y is label
Goal: Learn a function to map
x -> y
Examples: Classification, regression,
object detection, semantic
segmentation, image captioning, etc.
Unsupervised Learning
Data: x
Just data, no labels!
Goal: Learn some structure of
the data
Examples: Clustering, dimensionality
reduction, feature learning, generative
models, etc.
B.Sc. Final Project – Mohammad Pooya Malek
Traditional Neural Network, Decision tree, Nearest Neighbor
36
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Generative Adversarial Networks
B.Sc. Final Project – Mohammad Pooya Malek
37
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
38
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Yann LeCun Director of Facebook AI Research
“The most important breakthrough in deep learning, in my opinion, is
adversarial training (also called GAN for Generative Adversarial Networks).
This, and the variations that are now being proposed is the most interesting
idea in the last 10 years in ML.”
OpenAI
“In the process of training generative
models, we will endow the
computer with an understanding of
the world and what it is made up of.”
B.Sc. Final Project – Mohammad Pooya Malek
39
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Two-player game
Generator network: try to fool the discriminator by generating
real-looking images
Discriminator network: try to distinguish between real and fake
images
B.Sc. Final Project – Mohammad Pooya Malek
40
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
B.Sc. Final Project – Mohammad Pooya Malek
41
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
Generator
(Counterfeit)
B.Sc. Final Project – Mohammad Pooya Malek
42
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
Generator
(Counterfeit)
Discriminator
(Detective)
B.Sc. Final Project – Mohammad Pooya Malek
43
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
Generator
(Counterfeit)
Discriminator
(Detective)
Trained Data set
Training Set
B.Sc. Final Project – Mohammad Pooya Malek
44
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
CelebFaces Attributes Dataset
(CelebA)
more than 200K celebrity images,
each with 40 attribute annotations
10,177 number of identities,
202,599 number of face images, and
5 landmark locations, 40 binary
attributes annotations per image.
5_o_Clock_Shadow, Arched_Eyebrows, Attractive
Bags_Under_Eyes, Bald, Bangs, Big_Lips, Big_Nose, Black_Hai,r
Blond_Hair, Blurry Brown_Hair, Bushy_Eyebrow,s Chubby
Double_Chin, Eyeglasses, Goatee, Gray_Hair, Heavy_Makeup,
High_Cheekbones, Male, Mouth_Slightly_Open, Mustache,
Narrow_Eyes, No_Beard, Oval_Fac,e Pale_Skin Pointy_Nose
Receding_Hairline Rosy_Cheeks Sideburns, Smiling, Straight_Hair,
Wavy_Hair, Wearing_Earrings, Wearing_Hat, Wearing_Lipstick,
Wearing_Necklace, Wearing_Necktie,Young.
CelebA Dataset
B.Sc. Final Project – Mohammad Pooya Malek
45
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
MSCOCO Dataset
COCO-Stuff dataset: It
includes all 164K images
from COCO 2017 (train
118K, val 5K, test-dev
20K, test-challenge 20K).
It covers 172 classes: 80
thing classes, 91 stuff
classes and 1 class
'unlabeled'.
B.Sc. Final Project – Mohammad Pooya Malek
46
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
47
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
Generator
(Counterfeit)
Discriminator
(Detective)
Real
Fake
Trained Data set
Training Set
B.Sc. Final Project – Mohammad Pooya Malek
48
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
Generator
(Counterfeit)
Discriminator
(Detective)
Real
Fake
Trained Data set
(Training Set)
B.Sc. Final Project – Mohammad Pooya Malek
49
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
f
activations
gradients
“local gradient”
B.Sc. Final Project – Mohammad Pooya Malek
50
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
activations
Another example:
B.Sc. Final Project – Mohammad Pooya Malek
51
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
Generator
(Counterfeit)
Discriminator
(Detective)
Real
Fake
Trained Dataset
(Training Set)
B.Sc. Final Project – Mohammad Pooya Malek
52
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
GAN
Generative Adversarial Network
Input
Noise
Latent space coordinates
Generator
(Counterfeit)
Discriminator
(Detective)
Real
Fake
Trained Dataset
Training Set
B.Sc. Final Project – Mohammad Pooya Malek
53
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
54
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
55
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
56
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
VGGNet19 Architecture GAN
Recommended
Algorithm
B.Sc. Final Project – Mohammad Pooya Malek
57
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Transfer Learning with CNNs
1. Train on
ImageNet
2. If small dataset:
fix all weights
(treat CNN as fixed
feature extractor),
retrain only the
classifier
i.e. swap the
Softmax layer at
the end
3. If you have medium
sized dataset,
“finetune” instead: use
the old weights as
initialization, train the
full network or only
some of the higher
layers
retrain bigger portion
of the network, or even
all of it.
B.Sc. Final Project – Mohammad Pooya Malek
58
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Samples
B.Sc. Final Project – Mohammad Pooya Malek
Application of GAN in IRIB
Visual Effects (VFX) : GAN Can converts day to night theme
caused decrease cost of movies
Style transfer
Image Captioning
Image Inpainting
Robotics
,…
59
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Conclusion
B.Sc. Final Project – Mohammad Pooya Malek
60
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
References
 CS231n, Stanford University Deep learning Course
 MS COCO DataSet.
 CelebA DataSet.
 Criminisi A, Perez P, Toyama K. Region filling and object removal by exemplar-based
image inpainting[J]. IEEE Transactions on Image Processing, 2004, 13(9):1200-1212.
 Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley,
Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In
Advances in neural information processing systems, 2014.
B.Sc. Final Project – Mohammad Pooya Malek
61
Image Inpainting Using Deep Learning
Introduction | Related work | Problems | Algorithm | Samples | Conclusion
Way of Connecting to Author
MOPOMA1995@gmail.com
+989379595379
https://www.linkedin.com/in/mopoma1995/

More Related Content

What's hot

Image enhancement techniques
Image enhancement techniquesImage enhancement techniques
Image enhancement techniques
Saideep
 
Application of computer graphics and input devices
Application of computer graphics and input devicesApplication of computer graphics and input devices
Application of computer graphics and input devices
Mani Kanth
 

What's hot (20)

Color Image Processing
Color Image ProcessingColor Image Processing
Color Image Processing
 
Super resolution from a single image
Super resolution from a single imageSuper resolution from a single image
Super resolution from a single image
 
Matlab and Image Processing Workshop-SKERG
Matlab and Image Processing Workshop-SKERG Matlab and Image Processing Workshop-SKERG
Matlab and Image Processing Workshop-SKERG
 
Image processing second unit Notes
Image processing second unit NotesImage processing second unit Notes
Image processing second unit Notes
 
Halftoning in Computer Graphics
Halftoning  in Computer GraphicsHalftoning  in Computer Graphics
Halftoning in Computer Graphics
 
A Deep Journey into Super-resolution
A Deep Journey into Super-resolutionA Deep Journey into Super-resolution
A Deep Journey into Super-resolution
 
Image enhancement techniques
Image enhancement techniquesImage enhancement techniques
Image enhancement techniques
 
Super Resolution
Super ResolutionSuper Resolution
Super Resolution
 
Polygon filling
Polygon fillingPolygon filling
Polygon filling
 
Convolutional Neural Network (CNN)
Convolutional Neural Network (CNN)Convolutional Neural Network (CNN)
Convolutional Neural Network (CNN)
 
Image compression
Image compression Image compression
Image compression
 
COM2304: Intensity Transformation and Spatial Filtering – I (Intensity Transf...
COM2304: Intensity Transformation and Spatial Filtering – I (Intensity Transf...COM2304: Intensity Transformation and Spatial Filtering – I (Intensity Transf...
COM2304: Intensity Transformation and Spatial Filtering – I (Intensity Transf...
 
Matlab Working With Images
Matlab Working With ImagesMatlab Working With Images
Matlab Working With Images
 
Hit and-miss transform
Hit and-miss transformHit and-miss transform
Hit and-miss transform
 
Application of computer graphics and input devices
Application of computer graphics and input devicesApplication of computer graphics and input devices
Application of computer graphics and input devices
 
Morphological image processing
Morphological image processingMorphological image processing
Morphological image processing
 
Literature Review on Single Image Super Resolution
Literature Review on Single Image Super ResolutionLiterature Review on Single Image Super Resolution
Literature Review on Single Image Super Resolution
 
morphological image processing
morphological image processingmorphological image processing
morphological image processing
 
Chapter10 image segmentation
Chapter10 image segmentationChapter10 image segmentation
Chapter10 image segmentation
 
Color models in Digitel image processing
Color models in Digitel image processingColor models in Digitel image processing
Color models in Digitel image processing
 

Similar to Image Inpainting Using Deep Learning

Accelerating the Pace of Engineering Education with Simulation, Hardware and ...
Accelerating the Pace of Engineering Education with Simulation, Hardware and ...Accelerating the Pace of Engineering Education with Simulation, Hardware and ...
Accelerating the Pace of Engineering Education with Simulation, Hardware and ...
Joachim Schlosser
 
Developing Computational Skills in the Sciences with Matlab Webinar 2017
Developing Computational Skills in the Sciences with Matlab Webinar 2017Developing Computational Skills in the Sciences with Matlab Webinar 2017
Developing Computational Skills in the Sciences with Matlab Webinar 2017
SERC at Carleton College
 
deep_stereo_arxiv_2015
deep_stereo_arxiv_2015deep_stereo_arxiv_2015
deep_stereo_arxiv_2015
Ivan Neulander
 
Introduction to simulink (1)
Introduction to simulink (1)Introduction to simulink (1)
Introduction to simulink (1)
Memo Love
 
Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...
Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...
Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...
ESUG
 
Introduction to Machine Vision
Introduction to Machine VisionIntroduction to Machine Vision
Introduction to Machine Vision
Nasir Jumani
 

Similar to Image Inpainting Using Deep Learning (20)

NTU DBME5028 Week8 Transfer Learning
NTU DBME5028 Week8 Transfer LearningNTU DBME5028 Week8 Transfer Learning
NTU DBME5028 Week8 Transfer Learning
 
3.point operation and histogram based image enhancement
3.point operation and histogram based image enhancement3.point operation and histogram based image enhancement
3.point operation and histogram based image enhancement
 
Convex optmization in communications
Convex optmization in communicationsConvex optmization in communications
Convex optmization in communications
 
Week 9: Programming for Data Analysis
Week 9: Programming for Data AnalysisWeek 9: Programming for Data Analysis
Week 9: Programming for Data Analysis
 
Accelerating the Pace of Engineering Education with Simulation, Hardware and ...
Accelerating the Pace of Engineering Education with Simulation, Hardware and ...Accelerating the Pace of Engineering Education with Simulation, Hardware and ...
Accelerating the Pace of Engineering Education with Simulation, Hardware and ...
 
Developing Computational Skills in the Sciences with Matlab Webinar 2017
Developing Computational Skills in the Sciences with Matlab Webinar 2017Developing Computational Skills in the Sciences with Matlab Webinar 2017
Developing Computational Skills in the Sciences with Matlab Webinar 2017
 
deep_stereo_arxiv_2015
deep_stereo_arxiv_2015deep_stereo_arxiv_2015
deep_stereo_arxiv_2015
 
Log polar coordinates
Log polar coordinatesLog polar coordinates
Log polar coordinates
 
Overlapping community detection in Large-Scale Networks using BigCLAM model b...
Overlapping community detection in Large-Scale Networks using BigCLAM model b...Overlapping community detection in Large-Scale Networks using BigCLAM model b...
Overlapping community detection in Large-Scale Networks using BigCLAM model b...
 
Introduction to simulink (1)
Introduction to simulink (1)Introduction to simulink (1)
Introduction to simulink (1)
 
MATLAB for Researcher Course - ATIT Academy
MATLAB for Researcher  Course - ATIT AcademyMATLAB for Researcher  Course - ATIT Academy
MATLAB for Researcher Course - ATIT Academy
 
talk1
talk1talk1
talk1
 
Scilab Modelica conference 20150921
Scilab Modelica conference 20150921Scilab Modelica conference 20150921
Scilab Modelica conference 20150921
 
Modern Convolutional Neural Network techniques for image segmentation
Modern Convolutional Neural Network techniques for image segmentationModern Convolutional Neural Network techniques for image segmentation
Modern Convolutional Neural Network techniques for image segmentation
 
Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...
Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...
Towards Modularity in Live Visual Modeling: A case-study with OpenPonk and Ke...
 
Keynote at IWLS 2017
Keynote at IWLS 2017Keynote at IWLS 2017
Keynote at IWLS 2017
 
Lecture - Linear Programming.pdf
Lecture - Linear Programming.pdfLecture - Linear Programming.pdf
Lecture - Linear Programming.pdf
 
Mini Project- Torque Control of a DC Motor
Mini Project- Torque Control of a DC MotorMini Project- Torque Control of a DC Motor
Mini Project- Torque Control of a DC Motor
 
Introduction to Machine Vision
Introduction to Machine VisionIntroduction to Machine Vision
Introduction to Machine Vision
 
week 9 lec 15.pptx
week 9 lec 15.pptxweek 9 lec 15.pptx
week 9 lec 15.pptx
 

Recently uploaded

Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
9953056974 Low Rate Call Girls In Saket, Delhi NCR
 
notes on Evolution Of Analytic Scalability.ppt
notes on Evolution Of Analytic Scalability.pptnotes on Evolution Of Analytic Scalability.ppt
notes on Evolution Of Analytic Scalability.ppt
MsecMca
 
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort ServiceCall Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
9953056974 Low Rate Call Girls In Saket, Delhi NCR
 

Recently uploaded (20)

Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torque
 
Intze Overhead Water Tank Design by Working Stress - IS Method.pdf
Intze Overhead Water Tank  Design by Working Stress - IS Method.pdfIntze Overhead Water Tank  Design by Working Stress - IS Method.pdf
Intze Overhead Water Tank Design by Working Stress - IS Method.pdf
 
Thermal Engineering Unit - I & II . ppt
Thermal Engineering  Unit - I & II . pptThermal Engineering  Unit - I & II . ppt
Thermal Engineering Unit - I & II . ppt
 
Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
 
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
 
Thermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VThermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - V
 
notes on Evolution Of Analytic Scalability.ppt
notes on Evolution Of Analytic Scalability.pptnotes on Evolution Of Analytic Scalability.ppt
notes on Evolution Of Analytic Scalability.ppt
 
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
 
Top Rated Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...
Top Rated  Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...Top Rated  Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...
Top Rated Pune Call Girls Budhwar Peth ⟟ 6297143586 ⟟ Call Me For Genuine Se...
 
Unit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdfUnit 1 - Soil Classification and Compaction.pdf
Unit 1 - Soil Classification and Compaction.pdf
 
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdfONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
 
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
 
BSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptx
BSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptxBSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptx
BSides Seattle 2024 - Stopping Ethan Hunt From Taking Your Data.pptx
 
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
 
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...Booking open Available Pune Call Girls Koregaon Park  6297143586 Call Hot Ind...
Booking open Available Pune Call Girls Koregaon Park 6297143586 Call Hot Ind...
 
Call Girls Walvekar Nagar Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Walvekar Nagar Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Walvekar Nagar Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Walvekar Nagar Call Me 7737669865 Budget Friendly No Advance Booking
 
(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7
(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7
(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7
 
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort ServiceCall Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
 
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...Booking open Available Pune Call Girls Pargaon  6297143586 Call Hot Indian Gi...
Booking open Available Pune Call Girls Pargaon 6297143586 Call Hot Indian Gi...
 

Image Inpainting Using Deep Learning

  • 1. B.Sc. Final Project – Mohammad Pooya Malek 2 University of IRIB Faculty of Broadcast Engineering A Thesis Presented for the Degree of Bachelor of Science in Telecommunications Engineering Image Inpainting Based on Deep Learning By Mohammad Pooya Malek Supervisor Master Reviewer Dr. Azam Bastanfard Dr. Rahil Mahdian Summer 2018
  • 2. B.Sc. Final Project – Mohammad Pooya Malek 2 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Aim of Thesis:  Censorships Occluded Images Appropriate reconstruction To Complete the image which has been destroyed To repair old Images
  • 3. B.Sc. Final Project – Mohammad Pooya Malek Index 3 Introduction Related work Problems Recommended Algorithm (Proposed Model) Samples From Model Conclusion
  • 4. B.Sc. Final Project – Mohammad Pooya Malek Introduction 4 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 5. B.Sc. Final Project – Mohammad Pooya Malek Introduction 5 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Computer Vision (Machine Vision) • Artificial Intelligence (AI) • Deep Learning • Machine Learning
  • 6. B.Sc. Final Project – Mohammad Pooya Malek 6 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 7. B.Sc. Final Project – Mohammad Pooya Malek 7 Neural Network Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 8. B.Sc. Final Project – Mohammad Pooya Malek 8 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Neural Networks: Architectures “Fully-connected” layers “2-layer Neural Net”, or “1-hidden-layer Neural Net” “3-layer Neural Net”, or “2-hidden-layer Neural Net”
  • 9. B.Sc. Final Project – Mohammad Pooya Malek 9 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Neural Networks: Architectures
  • 10. B.Sc. Final Project – Mohammad Pooya Malek 10 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion 32 32 3 Convolution Layer 5x5x3 filter 32x32x3 image
  • 11. B.Sc. Final Project – Mohammad Pooya Malek 11 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion 32 32 3 Convolution Layer 32x32x3 image 5x5x3 filter convolve (slide) over all spatial locations Activation map 1 28 28
  • 12. B.Sc. Final Project – Mohammad Pooya Malek 12 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion 32 32 3 Convolution Layer 32x32x3 image 5x5x3 filter convolve (slide) over all spatial locations activation maps 2 28 28 consider a second, green filter 1
  • 13. B.Sc. Final Project – Mohammad Pooya Malek 13 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion 32 32 3 Convolution Layer activation maps 6 28 28 For example, if we had 6, 5x5 filters, we’ll get 6 separate activation maps: We stack these up to get a “new image” of size 28x28x6!
  • 14. B.Sc. Final Project – Mohammad Pooya Malek 14 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Preview: ConvNet is a sequence of Convolutional Layers, interspersed with activation functions 32 32 3 CONV, ReLU e.g. 6 5x5x3 filters 28 28 6 CONV, ReLU e.g. 10 5x5x6 filters CONV, ReLU …. 10 24 24
  • 15. B.Sc. Final Project – Mohammad Pooya Malek 15 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 16. B.Sc. Final Project – Mohammad Pooya Malek 16 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Convolutional Layer
  • 17. B.Sc. Final Project – Mohammad Pooya Malek 17 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Convolutional Layer http://setosa.io/ev/image-kernels/
  • 18. B.Sc. Final Project – Mohammad Pooya Malek 18 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Activation Functions Sigmoid tanh tanh(x) ReLU max(0,x) Leaky ReLU max(0.1x, x) ELU
  • 19. B.Sc. Final Project – Mohammad Pooya Malek 19 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Activation Functions ReLU (Rectified Linear Unit) - Computes f(x) = max(0,x) - Does not saturate (in +region) - Very computationally efficient - Converges much faster than sigmoid/tanh in practice (e.g. 6x) - Not zero-centered output - And the gradient when x < 0 is 0
  • 20. B.Sc. Final Project – Mohammad Pooya Malek 20 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Activation Functions Leaky ReLU - Does not saturate - Computationally efficient - Converges much faster than sigmoid/tanh in practice! (e.g. 6x) - will not “die”. Parametric Rectifier (PReLU) backprop into alpha (parameter)
  • 21. B.Sc. Final Project – Mohammad Pooya Malek 21 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Activation Functions Exponential Linear Units (ELU) - All benefits of ReLU - Does not die - Closer to zero mean outputs - Computation requires exp()
  • 22. B.Sc. Final Project – Mohammad Pooya Malek 22 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion ReLU (Rectified Linear Units) max(0,x)
  • 23. B.Sc. Final Project – Mohammad Pooya Malek 23 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Pooling layer
  • 24. B.Sc. Final Project – Mohammad Pooya Malek 24 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Fully Connected layer
  • 25. B.Sc. Final Project – Mohammad Pooya Malek 25 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion LeNet-5, a pioneering convolutional network by LeCun. in 1998 classifies digits Recognize hand-written numbers on 32x32 pixel images. LeNet Architecture
  • 26. B.Sc. Final Project – Mohammad Pooya Malek 26 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion AlexNet Architecture Alex Krizhevsky, Geoffrey Hinton, and Ilya Sutskever ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2012 Winner AlexNet uses ReLU(Rectified Linear Unit) for the non-linear part, instead of a Tanh or Sigmoid function The advantage of Using ReLU over sigmoid is that it trains much faster Full (simplified) AlexNet architecture: [227x227x3] INPUT [55x55x96] CONV1: 96 11x11 filters at stride 4, pad 0 [27x27x96] MAX POOL1: 3x3 filters at stride 2 [27x27x96] NORM1: Normalization layer [27x27x256] CONV2: 256 5x5 filters at stride 1, pad 2 [13x13x256] MAX POOL2: 3x3 filters at stride 2 [13x13x256] NORM2: Normalization layer [13x13x384] CONV3: 384 3x3 filters at stride 1, pad 1 [13x13x384] CONV4: 384 3x3 filters at stride 1, pad 1 [13x13x256] CONV5: 256 3x3 filters at stride 1, pad 1 [6x6x256] MAX POOL3: 3x3 filters at stride 2 [4096] FC6: 4096 neurons [4096] FC7: 4096 neurons [1000] FC8: 1000 neurons (class scores)
  • 27. B.Sc. Final Project – Mohammad Pooya Malek 27 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion VGGNet16 Architecture Visual Geometry Group Made improvement over AlexNet Increases the depth of the network which enable it to learn more complex features
  • 28. B.Sc. Final Project – Mohammad Pooya Malek 28 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion VGGnet19 Architecture
  • 29. B.Sc. Final Project – Mohammad Pooya Malek 29 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion CNN Architectures
  • 30. B.Sc. Final Project – Mohammad Pooya Malek 30 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Related work Criminisi Algorithm
  • 31. B.Sc. Final Project – Mohammad Pooya Malek 31 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 32. B.Sc. Final Project – Mohammad Pooya Malek 32 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Problems
  • 33. B.Sc. Final Project – Mohammad Pooya Malek 33 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Algorithm Supervised vs Unsupervised
  • 34. B.Sc. Final Project – Mohammad Pooya Malek 34 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion OpenCV (Open Source Computer Vision Library) BSD license: free for both academic and commercial usage Open Source: C++, Python and Java interfaces Cross-platform: Windows, Linux, Mac OS, iOS and Android. focus on real-time applications Originally developed by Intel 1999; supported by Willow Garage maintained by Itseez; Adopted all around the world: 47 thousand user and community downloads exceeding 14 million times; last release: 3.4.0 - 23 December 2017 Tensorflow BSD license: free for both academic and commercial usage Open Source: Python, C++, CUDA Cross-platform: Linux, macOS, Windows, Android, iOS and website. open source software library for high performance numerical computation TensorFlow was developed by the Google Brain team in 2011 Initial released on November 9, 2015; 2 years ago TensorFlow can run on multiple CPUs and GPUs last release: August 8, 2018; 34 days ago
  • 35. B.Sc. Final Project – Mohammad Pooya Malek 35 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Supervised vs Unsupervised Supervised Learning Data: (x, y) x is data, y is label Goal: Learn a function to map x -> y Examples: Classification, regression, object detection, semantic segmentation, image captioning, etc. Unsupervised Learning Data: x Just data, no labels! Goal: Learn some structure of the data Examples: Clustering, dimensionality reduction, feature learning, generative models, etc.
  • 36. B.Sc. Final Project – Mohammad Pooya Malek Traditional Neural Network, Decision tree, Nearest Neighbor 36 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Generative Adversarial Networks
  • 37. B.Sc. Final Project – Mohammad Pooya Malek 37 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 38. B.Sc. Final Project – Mohammad Pooya Malek 38 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Yann LeCun Director of Facebook AI Research “The most important breakthrough in deep learning, in my opinion, is adversarial training (also called GAN for Generative Adversarial Networks). This, and the variations that are now being proposed is the most interesting idea in the last 10 years in ML.” OpenAI “In the process of training generative models, we will endow the computer with an understanding of the world and what it is made up of.”
  • 39. B.Sc. Final Project – Mohammad Pooya Malek 39 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Two-player game Generator network: try to fool the discriminator by generating real-looking images Discriminator network: try to distinguish between real and fake images
  • 40. B.Sc. Final Project – Mohammad Pooya Malek 40 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates
  • 41. B.Sc. Final Project – Mohammad Pooya Malek 41 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates Generator (Counterfeit)
  • 42. B.Sc. Final Project – Mohammad Pooya Malek 42 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates Generator (Counterfeit) Discriminator (Detective)
  • 43. B.Sc. Final Project – Mohammad Pooya Malek 43 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates Generator (Counterfeit) Discriminator (Detective) Trained Data set Training Set
  • 44. B.Sc. Final Project – Mohammad Pooya Malek 44 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion CelebFaces Attributes Dataset (CelebA) more than 200K celebrity images, each with 40 attribute annotations 10,177 number of identities, 202,599 number of face images, and 5 landmark locations, 40 binary attributes annotations per image. 5_o_Clock_Shadow, Arched_Eyebrows, Attractive Bags_Under_Eyes, Bald, Bangs, Big_Lips, Big_Nose, Black_Hai,r Blond_Hair, Blurry Brown_Hair, Bushy_Eyebrow,s Chubby Double_Chin, Eyeglasses, Goatee, Gray_Hair, Heavy_Makeup, High_Cheekbones, Male, Mouth_Slightly_Open, Mustache, Narrow_Eyes, No_Beard, Oval_Fac,e Pale_Skin Pointy_Nose Receding_Hairline Rosy_Cheeks Sideburns, Smiling, Straight_Hair, Wavy_Hair, Wearing_Earrings, Wearing_Hat, Wearing_Lipstick, Wearing_Necklace, Wearing_Necktie,Young. CelebA Dataset
  • 45. B.Sc. Final Project – Mohammad Pooya Malek 45 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion MSCOCO Dataset COCO-Stuff dataset: It includes all 164K images from COCO 2017 (train 118K, val 5K, test-dev 20K, test-challenge 20K). It covers 172 classes: 80 thing classes, 91 stuff classes and 1 class 'unlabeled'.
  • 46. B.Sc. Final Project – Mohammad Pooya Malek 46 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 47. B.Sc. Final Project – Mohammad Pooya Malek 47 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates Generator (Counterfeit) Discriminator (Detective) Real Fake Trained Data set Training Set
  • 48. B.Sc. Final Project – Mohammad Pooya Malek 48 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates Generator (Counterfeit) Discriminator (Detective) Real Fake Trained Data set (Training Set)
  • 49. B.Sc. Final Project – Mohammad Pooya Malek 49 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion f activations gradients “local gradient”
  • 50. B.Sc. Final Project – Mohammad Pooya Malek 50 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion activations Another example:
  • 51. B.Sc. Final Project – Mohammad Pooya Malek 51 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates Generator (Counterfeit) Discriminator (Detective) Real Fake Trained Dataset (Training Set)
  • 52. B.Sc. Final Project – Mohammad Pooya Malek 52 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion GAN Generative Adversarial Network Input Noise Latent space coordinates Generator (Counterfeit) Discriminator (Detective) Real Fake Trained Dataset Training Set
  • 53. B.Sc. Final Project – Mohammad Pooya Malek 53 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 54. B.Sc. Final Project – Mohammad Pooya Malek 54 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 55. B.Sc. Final Project – Mohammad Pooya Malek 55 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion
  • 56. B.Sc. Final Project – Mohammad Pooya Malek 56 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion VGGNet19 Architecture GAN Recommended Algorithm
  • 57. B.Sc. Final Project – Mohammad Pooya Malek 57 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Transfer Learning with CNNs 1. Train on ImageNet 2. If small dataset: fix all weights (treat CNN as fixed feature extractor), retrain only the classifier i.e. swap the Softmax layer at the end 3. If you have medium sized dataset, “finetune” instead: use the old weights as initialization, train the full network or only some of the higher layers retrain bigger portion of the network, or even all of it.
  • 58. B.Sc. Final Project – Mohammad Pooya Malek 58 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Samples
  • 59. B.Sc. Final Project – Mohammad Pooya Malek Application of GAN in IRIB Visual Effects (VFX) : GAN Can converts day to night theme caused decrease cost of movies Style transfer Image Captioning Image Inpainting Robotics ,… 59 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Conclusion
  • 60. B.Sc. Final Project – Mohammad Pooya Malek 60 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion References  CS231n, Stanford University Deep learning Course  MS COCO DataSet.  CelebA DataSet.  Criminisi A, Perez P, Toyama K. Region filling and object removal by exemplar-based image inpainting[J]. IEEE Transactions on Image Processing, 2004, 13(9):1200-1212.  Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In Advances in neural information processing systems, 2014.
  • 61. B.Sc. Final Project – Mohammad Pooya Malek 61 Image Inpainting Using Deep Learning Introduction | Related work | Problems | Algorithm | Samples | Conclusion Way of Connecting to Author MOPOMA1995@gmail.com +989379595379 https://www.linkedin.com/in/mopoma1995/