5cb9a0bcfd72dfb20d947187dff978b5 (1).pdf

### A single paragraph on AI

**Artificial Intelligence (AI)** is a transformative field of computer science dedicated to creating systems that can perform tasks typically requiring human intelligence. This includes everything from simple problem-solving to complex abilities like learning, reasoning, and language understanding. At its core, AI relies on powerful algorithms and vast amounts of data to identify patterns and make predictions. Key sub-fields, such as **Machine Learning** and **Deep Learning**, have revolutionized technology, enabling applications like personalized recommendations, self-driving cars, and virtual assistants. As AI continues to evolve, it promises to reshape industries and redefine our daily lives, while also presenting crucial ethical challenges related to data privacy, bias, and its impact on the workforce.

1. Introduction to Computer Vision
PrinsipkecerdasanArtifisial
Dr. DyahArumingTyas.
1

2. What is Computer Vision ?

6. What is Computer Vision?
Make computers understand images and videos.
•What kind of scene?
•Where are the cars?
•How far is the building?

7. What is Computer Vision?
Make computers understand images and videos.
•What are they doing?
•Why is this happening?
•What is important?
•What will I see?

8. The goal of computer vision is to give
computers (super) human-level perception

9. typical perception pipeline

10. typical perception pipeline

11. typical perception pipeline

12. typical perception pipeline

13. Important note: In general, computer vision
does not work

14. Important note: In general, computer vision
does not work
(except in certain situation / condition)

15. Computer Vision
•Computer Vision
15
Computer
Science
Mathematics
Engineering
Physics
Biology
Psychology
Computer
Vision
Neuroscience
Cognitive
Sciences
graphics,
algorithms,
theory
systems,
architecture
Information
Retrieval
Machine Learning
Robotics
Natural Language
Processing
Speech
Optics

16. Computer Vision
•Com
16
Computer
Artificia
nVisio
puter Vision
Machine
Learning
l Intelligence
Natural
Language
Processing
Robotics
Deep Learning

17. Computer Vision
•One of the fastest growing field of A.I.
• Computer Vision
17

18. Computer Vision
•Cisco: by 2016, >85% of Internet data is in form of pixels (multimedia)
• Computer Vision
19

19. Computer Vision
•YouTube :
every second,
about 10 hours of videos
are being uploaded
•Impossible to process
manually
http://www.everysecond.io/youtube
• Computer Vision
20

20. Age of Vision
21
• Computer Vision

21. Not This Vision
• Computer Vision
22

22. Computer Vision
•Vision data (raw pixel data) are one of the hardest data to harness
•Digital Dark Matter of The Internet
• Computer Vision
23

23. Where’s Waldo?
24
•Computer Vision

24. Where’s Waldo? (Extreme Edition)
•Computer Vision
25

25. Where’s The Real-life Waldo?
26
•Computer Vision

26. 27
•Computer Vision

27. What is Vision?
•Computer Vision
28

28. Hubel & Wiesel, 1959
•Computer Vision
34

29. Hubel & Wiesel’s Cat
•Computer Vision
35

30. Visual Perception
•Computer Vision
37

31. •Computer Vision
38

32. How to create good representations
of the visual world in a way it could be used to support
recognition?
39

33. Block World
[Larry Roberts, 1963]
•Edges define the world
•Extract edge-like Structures
•The beginning of Computer Vision
1963
1959
CV
• Computer Vision
40

34. The Birth of Computer Vision
Marvin Minsky and Seymour Papert
1963
1959
1966
CV
• Computer Vision
41

35. Stages of Visual Representation
[David Marr, 1970s]
1963
1959
1966
1970
CV
• Computer Vision
42

36. Recognition via Edge Detection
•Computer Vision
43
[John Canny, 1986]
[David Lowe, 1987] 1963
1959
CV
1966
1970
1986
AI Winter II

37. Recognition via Grouping
Normalized Cuts
[Shi & Malik, 1997]
CV AI Winter II
1959 1997
1963
1966
1970
1986
• Computer Vision
44

38. Recognition via Matching
Scale-Invariant Feature Transform (SIFT)
[David Lowe, 1999]
1963
1959
1966
1970
CV
1986
1997
AI Winter II
1999
• Computer Vision
45

39. Face Detection
[Viola & Jones, 2001]
CV AI Winter II
1959 1997
1963 1999
1966 2001
1970
1986
• Computer Vision
46

40. Recognition via Features and Parts
Deformable Part Model
[Felzenswalb, McAllester,
Ramanan, 2009]
1963
1959
1966
1970
CV
1986
1997
AI Winter II
1999
2001
2005
Histogram of Oriented
Gradients (HOG)
[Dalal & Triggs, 2005]
• Computer Vision
47
Popularity of
AI/ML in CV

41. There is a number of visual recognition problems that are related to image classification,
such as
object detection, image captioning, video classification
56
• Computer Vision

42. Basic Computer Vision Challenge
• Computer Vision
57

43. Basic Computer Vision Challenge
• Computer Vision
58

44. The quest for visual intelligence
goes far beyond object recognition…
• Computer Vision
59

45. CVPR Trends
60

46. CVPR Trends
61

47. CVPR Trends
62

48. The Quest of Visual Intelligence
• Computer Vision
63

49. Why is Computer Vision Hard?

50. Why is Computer Vision Hard?

51. Why is Computer Vision Hard?

52. Why is Computer Vision Hard?

53. Why is Computer Vision Hard?

54. Why is Computer Vision Hard?

55. Computer: okay, it’s a funny picture

56. Computer Vision Technology
Can Better Our Lives
• Computer Vision
71

57. Machine Vision

58. Object Recognition

59. Face Detection

60. Vision in Car

62. Computer Vision AR

64. Break, Question?
75
•Computer Vision

65. Where’s Waldo? (Extreme Edition)
•Computer Vision
82

66. Image Classification
3
•Computer Vision

67. Image Classification
Classifier cat
Train
8
4
…

68. Classic
Feature Extractions
model
training
c-Class scores
Deep Learning
End-to-end learning
model
training
c-Class scores
• Computer Vision
85

69. Computer Vision Core
•Computer Vision
Image Classification
Objects DetectionObjects Annotation
Object Localization
Image Captioning
Instance
Segmentation
Identification
Content-based
Image Retrieval
Character
Recognition
8
6

70. Recognition and Detection
8
7

71. Computer Vision Tasks
Classification
Classification
+ Localization Object Detection
Instance
Segmentation
Single Object Multiple Objects
CAT CAT
�, �, �, ℎ
CAT , DOG , DUCK
�, �, �, ℎ , �, �, �, ℎ , �, �, �, ℎ
3/8/2023
• Computer Vision
7
CAT, DOG, DUCK

72. Classification
•Input
•Output
: Image,
there may be an object inside
: Class Label
Single object per image
•Evaluation metric:
Accuracy
•Also called Recognition, Identification
CAT
• Computer Vision
3/8/2023 89

73. Localization
•Input
•Output
: Image,
there IS an object inside
: Box in the image (�, �, �, ℎ)
Single object, specific
•Evaluation metric:
Intersection over Union
(�, �, �, ℎ)
• Computer Vision
3/8/2023 90

74. Classification + Localization
•Input: Image,
there may be an object inside
•Output : Class Label and Box (�, �, �, ℎ)
•Evaluation metric:
Intersection over Union
Accuracy
•Only one object, simpler than detection
•Single Object Detection
CAT
• Computer Vision
3/8/2023 91
(�, �, �, ℎ)

75. Object Detection
•Input: Image, there may be
one or more objects inside
•Output : Class Label and Box (�, �, �, ℎ)
for each object
•Evaluation metric:
Intersection over Union
Accuracy, MAP, AR
CAT , DOG , DUCK
�, �, �, ℎ , �, �, �, ℎ , �, �, �, ℎ
• Computer Vision
3/8/2023 92

76. Semantic Segmentation
•Input: Image, there may be
one or more objects inside
•Output : Pixel label
•Evaluation metric:
Intersection over Union
Jaccard Index, Dice,
Accuracy, MAP, AR
• Computer Vision
3/8/2023 93

77. Instance Segmentation
•Input: Image, there may be
one or more objects inside
•Output : Pixel label for each instance
•Evaluation metric:
Intersection over Union
Jaccard Index, Dice,
Accuracy, MAP, AR
• Computer Vision
3/8/2023 94

78. Summing it up
• Computer Vision
3/8/2023 95
•Classification / Recognition
•Localization
•Classification + Localization
•Object Detection
•Segmentation
•Instance Segmentation
: What is
: Where is
: What and where is
: What and where are
: Which label pixel is
: Which instance pixel is

79. TerimaKasih