9/20/2023

Computer Vision
With Deep
Learning
Lecture 1: Introduction to Computer Vision

1. What’s Computer vision?

2. Image Representation

3. Color Spaces

Content 4. Image Transformations

5. Filters and Convolutions

6. Useful Libraries

7. Exercise 1: Loading and Pre-processing Images with

TensorFlow

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 2

1
 9/20/2023

1. What’s Computer
vision?

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 3

What’s Computer Vision?

❑ “At an abstract level, the goal of computer
vision problems is to use the observed image
data to infer something about the world”.
Page 83, Computer Vision: Models,
Learning, and Inference, 2012.

❑ Computer vision is a field of artificial

intelligence (AI) that enables computers to
derive meaningful information from digital
images and videos and understand the
visual world.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 4

2
 9/20/2023

Related Domains
❑Signal Processing: process electronic signals to
remove noise, extract information or prepare
them for further processing.
❑Image Processing: process digital images to
improve their quality, convert them into another
format or change them for further processing.
❑Machine Learning: help computers to
recognize patterns in images and understand
what they see.
❑Robot Vision and Machine Vision are
applications of computer vision in Robotics and
Industrial domains.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 5

Robot Vision

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 6

3
 9/20/2023

Machine Vision

Automatic Industrial Inspection

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 7

Computer Vision Tasks

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 8

4
 9/20/2023

Advanced Computer Vision Tasks

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 9

Computer Vision Applications

❑ Computer Vision has a many applications such as in retail, security, sports, healthcare,
automotive, manufacturing, and agriculture.
ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 10

10

5
 9/20/2023

Computer Vision in Retail

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 11

11

Computer Vision in Security

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 12

12

6
 9/20/2023

Computer Vision in Sports

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 13

13

Computer Vision in Healthcare

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 14

14

7
 9/20/2023

Computer Vision in Automotive

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 15

15

Computer Vision in Manufacturing

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 16

16

8
 9/20/2023

Computer Vision in Agriculture

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 17

17

2. Image Representation

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 18

18

9
 9/20/2023

Grayscale Image Representation

A grayscale image

0 = black
2D matrix of pixels 255 = White

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 19

19

RGB Image Representation

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 20

20

10
 9/20/2023

Pixel Grid System

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 21

21

Image Coordinates

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 22

22

11
 9/20/2023

3. Color Spaces

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 23

23

Color Spaces: Gray, RGB and HSV

In general, 3 color spaces are prevalent in modern
computer vision:

❑ GRAY: It eliminates color information by translating

to shades of gray: Y = 0.299xR + 0.587xG + 0.114xB.

❑ RGB: Each pixel is a three-element array, each value

representing the red, green, and blue colors.

❑ HSV: hue is a color tone, saturation is the intensity

of a color, and value represents its brightness.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 24

24

12
 9/20/2023

Convert RGB to Grayscale

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 25

25

Convert RGB to HSV

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 26

26

13
 9/20/2023

4. Image Transformations

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 27

27

Image Resizing
❑ Various interpolation algorithms can be used to interpolate the new pixel values e.g., linear,
cubic, and nearest neighbor.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 28

28

14
 9/20/2023

Image Cropping
❑ Cropping an image to get the region of interest (ROI) can be done easily by slicing the image
array and taking the new indices values.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 29

29

Geometric Transformation

Original Image

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 30

30

15
 9/20/2023

Thresholding
❑ In image processing, thresholding means to update the pixel’s value to either white or black
according to a threshold value.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 31

31

5. Filtering & Convolution

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 32

32

16
 9/20/2023

Image Filtering
❑ Image filtering is a fundamental concept in image
processing.

❑ Image filtering involves modifying each pixel's

value in an image based on the values of its
neighboring pixels.

❑ A filter (or kernel) is applied to an image to:

▪ Improve image quality

▪ Highlight specific details

▪ Remove noise and artifacts

▪ Detect edges and patterns

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 33

33

Types of Image Filters

❑ Smoothing Filters (Low-Pass Filters)
▪ Examples: Gaussian, Mean, Median filters.

▪ Used for noise reduction and blurring.

❑ Sharpening Filters (High-Pass Filters)

▪ Examples: Laplacian, Sobel, Prewitt filters.

▪ Enhance edges and fine details.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 34

34

17
 9/20/2023

Image Convolution
❑ Image convolution is a mathematical
operation that enhances, modifies, or extracts
features in an image by applying a filter or
kernel based on the values of neighboring
pixels.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 35

35

2D Convolution

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 36

36

18
 9/20/2023

6. Useful Libraries

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 37

37

OpenCV

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 38

38

19
 9/20/2023

TensorFlow
❑ End-to-end open-source platform
for machine learning and artificial
intelligence.

❑ TensorFlow is used for the training

and inference of deep neural
networks.

❑ Multi OS (Windows, Linux, Android,

Mac OS) and multi processor (CPU,
GPU, TPU, Cloud).
ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 39

39

❑ Deep learning API written in Python and running on top of the machine learning

platform TensorFlow.

❑ User-friendly that is designed to enable easy and fast experimentation with DL.

❑ Reliable, modular, and extensible that works with multi-backend and multi-platform.

❑ Powerful that provides industry-strength performance and scalability.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 40

40

20
 9/20/2023

Installation
❑ Step (1): Download the *.exe file from https://www.anaconda.com/download/ and
install Anaconda.

❑ Step (2): Lunch Anaconda’s CMD terminal and execute ‘conda create --name myenv’
to create a new environment myenv and active using ‘conda activate myenv’.

❑ Step (3): Install Tensorflow using ‘conda install -c conda-forge tensorflow-gpu’ for
NVIDA Graphic cards.

❑ Step (4): Install OpenCV using ‘conda install -c conda-forge opencv’.

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 41

41

EX1: Loading and Pre-processing Images

with TensorFlow

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 42

42

21
 9/20/2023

Loading Built-in Image Dataset

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 43

43

Image
Visualization

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 44

44

22
 9/20/2023

Image
Preprocessing

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 45

45

Image Transformation

ISITCOM 3DNI OLFA.BESBES@ISITC.U-SOUSSE.TN 2023-2024 46

46

23