Convolutional
Neural
Networks
Deep Learning
IndabaXCentrafriquetheitronics.com
Tsopnang SR
Content:
Introduction
What is CNN?
Why CNN?
Practical Aspect of CNN
Demo 1: Image Classifer
Project 1: Feed our pets
Project 2: Smart House
Tsopnang SR
Introduction
CNN is here!!!
Tsopnang SR
What is CNN? (1)
CNN is here!!!
Tsopnang SR
What is CNN? (2)
CNN is here!!!
Tsopnang SR
What is CNN? (3)
CNN is here!!!
Tsopnang SR
Why CNN?
CNN is here!!!
Tsopnang SR
Computer Vision Problems
Image Classifcation
Cat? (0/1)
Neural Style Transfer
Object detection
64x64
Tsopnang SR
Deep Learning on large images
Cat? (0/1)
 
 
 
 
    
64x64
1000x1000x3
= 3 millions
W[1] = (1000x3M)
= 3 billions
Tsopnang SR
Computer Vision Problem
vertical edges
horizontal edges
Tsopnang SR
Vertical edge detection
3 0 1 2 7 4
1 5 8 9 3 1
2 7 2 5 1 3
0 1 3 1 7 8
4 2 1 6 2 8
2 4 5 2 3 9
∗   
1
1
1
-1
-1
-1
0
0
0
1
1
1
-1
-1
-1
0
0
0
Tsopnang SR
Vertical edge detection
10 10 10 0 0 0
10 10 10 0 0 0
10 10 10 0 0 0
10 10 10 0 0 0
10 10 10 0 0 0
10 10 10 0 0 0
1 0 -1
1 0 -1
1 0 -1
 
0 30 30 0
0 30 30 0
0 30 30 0
0 30 30 0
 
Tsopnang SR
Padding
∗  
Tsopnang SR
Padding
∗  
Tsopnang SR
Strided convolution
flter
 
image
 
padding p stride s
 
× 
 
Tsopnang SR
Valid and Same convolutions
“Valid”: Output: n – f + 1 * n – f +1
“Same”: Pad so that output size is the
same as the input size.
Output: n – f + 2p + 1 * n – f + 2p +1
Tsopnang SR
Convolutions on RGB image
  
4 x 4
6 x 6 x 3
4 x 4
3 x 3
Tsopnang SR
Multiple flters
 
6 x 6 x 3
4 x 4
3 x 3 x 3
∗ 
∗ 
3 x 3 x 3
 
4 x 4
4 x 4 x 2
Tsopnang SR
Project 1: Let's Feed our pets
Tsopnang SR
Project 1: Let's Feed our pets
Tsopnang SR
Project 2: Smart House
Tsopnang SR
Project 2: Smart House
Tsopnang SR
Project 2: Smart House
http://theitronics.com/blog
/Smart-House-And-Factory
Thanks for your
kind attention!
Tel: +237696307380
Email: tsopnangsr@gmail.com
theitronics.com

Demo3 Convolutional Neural Network