See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/341526917

Computer Vision versus Human Vision

Presentation · April 2020

DOI: 10.13140/RG.2.2.33384.90880

CITATION READS

1 1,190

1 author:

Professor G R Sinha
International Institute of Information Technology Bangalore
314 PUBLICATIONS 2,048 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Computational Intelligence and Applications for Pandemics and Healthcare View project

Reliable CAD system for Breast cancer detection View project

All content following this page was uploaded by Professor G R Sinha on 20 May 2020.

The user has requested enhancement of the downloaded file.

Computer Vision versus Human Vision

G R Sinha, PhD
IEEE Senior Member, ACM Distinguished Speaker, IEEE Distinguished Speaker
Adjunct Professor IIIT Bangalore & Professor, Myanmar Institute of Information Technology Mandalay
Email: [email protected], [email protected], [email protected]
 Lecture Outline

 Background

 Human Vision

 Computer Vision

 Computer Vision Versus Human Vision

 Few Applications

 Concluding Remarks

Source: Google Image

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 2
 About ACM
 ACM, the Association for Computing Machinery (www.acm.org), is the premier global community of computing professionals
and students with nearly 100,000 members in more than 170 countries interacting with more than 2 million computing
professionals worldwide.

 OUR MISSION: We help computing professionals to be their best and most creative. We connect them to their peers, to what
the latest developments, and inspire them to advance the profession and make a positive impact on society.

 OUR VISION: We see a world where computing helps solve tomorrow’s problems – where we use our knowledge and skills to
advance the computing profession and make a positive social impact throughout the world.

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 The Distinguished Speakers Program
is made possible by

For additional information, please visit http://dsp.acm.org/

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Background

5
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Human & Computer Vison

 Human Vision (HV) has Eyes and Computer Vision (CV) has Sensors

 HV understands stimuli and CV understands numbers

https://www.google.co.in/search?q=computer+vision&tbm=isch&ved=2ahUKEwi77vOvmsbnAhXckEsFHYTzBewQ2-
cCegQIABAA&oq=computer+&gs_l=img.1.0.0i67l9j0.39160.41186..42710...0.0..0.233.1046.3j5j1......0....1..gws-wiz-img.sFoN2TlO37g&ei=7uNAXru_OtyhrtoPhOeX4A4&bih=625&biw=1366#imgrc=MC8VVzWkpg630M

6
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 CV involves machine

 CV makes computers understand images and video

 A picture is worth more than 10000 words.

 What kind of scene?

 Where are the cars?

 How far is the building?

 …

Source: Google Images

7
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Why CV matters?

Safety Health Security

Comfort Fun Access

* Source: Google Images

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 8
 A Rough History*

 1966: Marvin Minsky of AI & MIT lab fame gives Gerald Jay Sussman, an undergraduate student
a summer project to link a computer to a camera and get the computer to “describe what it
saw”. Thus, it begins the quest for solving the “visual input” problem which was then considered
easier than cognitive intelligence problem in machines. It continues to baffle the experts to date.

 1970’s: Some progress on interpreting selected images

 1980’s: ANNs come and go; shift toward geometry and increased mathematical rigor

 1990’s: Face recognition; statistical analysis

 2000’s: Broader recognition; large annotated datasets available; video processing starts

* Source: Wikipedia and https://medium.com/analytics-vidhya/computer-vision-from-a-summer-intern-project-to-redefining-ai-future-5dc87fdc9f72

9
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Human Perception has shortcomings

https://www.google.co.in/search?q=Sinha+and+Poggio,+Nature,+1996&hl=en-
US&source=lnms&tbm=isch&sa=X&ved=2ahUKEwj6qaCbncbnAhWtyjgGHUinBxQQ_AUoAXoECAwQAw&biw=
1366&bih=625#imgrc=qcV89bpSmAv2aM

10
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Interesting Story

“How the Afghan Girl was Identified by Her Iris Patterns”: https://en.wikipedia.org/wiki/Afghan_Girl
& https://www.youtube.com/watch?v=cK6EnFu3NHc

11
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Addressing Uncertainty

https://www.google.co.in/search?q=Smile+Detection&tbm=isch&source=iu&ictx=1&fir=9fl-gFtHC27aUM%253A%252C1-pUThf5NCWheM%252C_&vet=1&usg=AI4_-
kRJlYDHACM50Ovm3HgYbStSdS84YQ&sa=X&ved=2ahUKEwitsaSIssbnAhXdxDgGHTVYCqcQ_h0wEnoECAsQCA&biw=1366&bih=625#imgrc=9fl-gFtHC27aUM:

12
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Mobile Robots & Vyom Mitra

https://www.google.co.in/search?q=mobile+robots+in+isro+for+space+mission&tbm=isch&ved=2ahUKEwif_qjG3sjnAhX1XHwKHbNuDs8Q2-
cCegQIABAA&oq=mobile+robots+in+isro+for+space+mission&gs_l=img.3...37305.38793..40133...1.0..0.95.455.5......0....1..gws-wiz-img.3sPQn-DNb8w&ei=2zdCXt-
kA_W58QOz3bn4DA&bih=576&biw=1366#imgrc=oEa8yUEUaN00RM

13
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Human Vision

14
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Human Eye

 Human eye is a complete imaging system.

Ciliary Muscle
Sclera Ear side (Temporal)
Iris Vitreous Humor

Fovea
Pupil Eyelens

Retina

Optic Nerve
Cornea
Nose side (Nasal)
Aqueous Humor

Choroid
Suspensory ligament

15
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Human Visual System
 Image Formation: https://www.youtube.com/watch?v=5AJ0YOwl55A

Exposure
Image formation Detection Processing
Control

 Cornea  Iris/pupil  Retina  Brain

 Lens  Photoreceptor  Rods
 sensitivity  Cones

16
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 contd..

 Curved surfaces of the eye focus the image onto the back surface of the eye.

Object Image

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 17
 Cornea

 Outer wall of the eye is formed by the hard, white sclera.

 Cornea is the clear portion of the sclera.

 2/3 of the refraction takes place at the cornea.

Sclera

Cornea

18
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Iris and Pupil

 Colored iris controls the size of the opening (pupil) where the light enters.
 Pupil determines the amount of light, like the aperture of a camera.
Iris

Pupil

Iris open Iris closed

Dilated pupil Constricted pupil

19
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Lens

Ciliary muscle

Lens
 Eye lens is made of transparent fibers in a clear membrane,

suspended by suspensory ligament

Suspensory
Ligament
 Used as a fine focusing mechanism by the eye; provides 1/3 of
Transparent
eye’s total refracting power
Fibers

 Non-uniform index of refraction

Cross section of the eye lens

20
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Retina

 Retina is the photosensitive “detector” for the eye.

 Two types of receptors in the retina: rods for low light level, and cones for color.

 Fovea, located at the center of the retina, contains a greater concentration of cones.

 Signals from the receptors leave through the optic nerve to the brain.

Retina

Fovea

Optic Nerve

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 21
 contd..
 Retina is made of network of nerve cells.
• Network works together to reduce the amount of information in a process called lateral inhibition.
Light Cones

Rods

To optic nerve Bipolar

cells
Amicrine
cells
Ganglion
cells
Horizontal
cells

22
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Retina Layers

 Retina is made of three layers:

Fovea  Plexiform layer is a network of nerves which

carry the signals from the photo receptors.
Photo receptors

Light  Photo receptors.

Plexiform Layer
 Choroid provides nourishment to the
receptors, as well as absorb any light that
didn’t get absorbed by the photo receptors.

Optic Nerve Choroid

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 23
 Rods and Cones

Synaptic endings
Cell nucleus
Inner segments

Outer segments

Rod Cone
 Highly sensitive to low light level or  Sensitive to high light level or photopic
scotopic conditions. conditions.
 Black and white.  Three types of cones responsible for color
 Dispersed in the periphery of the retina. vision.
 Concentrated in the fovea.

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 24
 Adaptation in HV

 Why can’t you see immediately after you enter a movie theater from daylight?

 The threshold of detection changes with overall light level.

 The switch is quite gradual, until the sensitivities of cones and rods cross over at about 7 minutes in
the dark.
Photopic (cones)

Threshold of detection
(log scale)
Scotopic (rods)

0 5 10 15 20 25 30
Time in dark (minutes)

Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 25
 HV Typical Flow

Exposure Image
Control Processing
• Cornea • Retina
• Lens • Iris • Rods
•Brain
• Pupil • Cones
Image Image
Formation Detection

26
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Computer Vision (CV)

27
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 CV

 Computer vision is an interdisciplinary scientific field that deals with how computers can gain high-
level understanding from digital images or videos. From the perspective of engineering, it seeks to
understand and automate tasks that the human visual system can
do(https://en.wikipedia.org/wiki/Computer_vision)

 To simplify Computer vision (CV) can be described as: the use of machine language, deep learning
algorithms; to analyze, infer, interpret and understand the content of images or videos; and to
derive information (https://medium.com/analytics-vidhya/computer-vision-from-a-summer-intern-project-to-
redefining-ai-future-5dc87fdc9f72).

28
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 contd..

48 68 89 108 123 138 149 158 164 167

49 70 91 109 124 139 150 158 167 165

51 71 91 109 124 139 151 159 165 168

52 70 91 110 125 140 151 159 165 168

52 71 92 110 126 140 151 159 166 168

53 72 92 110 126 140 151 159 166 168

53 73 93 111 126 140 151 159 166 168

54 74 93 111 126 140 151 159 166 168

54 74 93 110 126 140 151 160 166 168

55 75 93 111 126 140 151 160 166 168

29
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Digital Image and its Processing

 Images are typically generated by illuminating a scene and absorbing the energy reflected by the
objects in that scene.

Source: DIP by Gonzalez & Woods

30
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 contd..

 Digital Image - f ( x, y ) a two-dimensional function with x and y are spatial coordinates and the

amplitude of f is called intensity or gray level at the point (x, y)

 Digital Image Processing:

1. Low-level: inputs and outputs are images

2. Mid-level: outputs are attributes extracted from input images

3. High-level: an ensemble of recognition of individual objects

31
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 (contd..)

Low Level Processing Mid Level Processing High Level Processing

Input: Image Input: Image Input: Attributes
Output: Image Output: Attributes Output: Understanding
Examples: Noise Examples: Object Examples: Scene
removal, image recognition, understanding,
sharpening segmentation autonomous navigation

32
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Major Steps

Image acquisition Image Processing Understanding

Description
Image Segmentation
Image Restoration

Image Enhancement

Image Acquisition

33
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 CV versus HV

34
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 HV vs CV

Source: Google Images

35
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Privacy and Field of View

 Privacy: Computer vision searches and analyzes countless images and videos, and other pieces of
information, which are stored online somewhere.

 Relative Field of View: Human field of view is 220 degrees and CV can have 360 degree field of view.

Source: Google Images

36
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 How far we can see

 Standing on the flat surface with our eyes about 5 feet off the ground, we can see the objects 3
miles away.
 Detecting a candle flame Researchers believe that without obstructions, a person with healthy but
average vision could see a candle flame from as far as 1.6 miles.
 Depending on the resolution, sensor and lens of the camera, it can see nearly 0.14 mile only.

37
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Persistence
 Persistence of vision of normal human eye is 1/16s.
 Human eye and brain can only process 10 to 12 separate images per second.
 Persistence doesn’t matter in CV.

Source: Google Images

38
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Both are biased

 “Debiasing humans is harder than debiasing AI systems.” OLGA RUSSAKOVSKY, PRINCETON

(https://www.wired.com/story/ai-biased-how-scientists-trying-fix/)

39
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Few Applications

40
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Smart City and Smart Transportation System

https://www.google.co.in/search?q=computer+vision+based+car&source=lnms&tbm=isch&sa=X&ved=2ahUKEwjTu46u28jnAhWAxzgGHZqgBcgQ_AUoAXoECA4QAw&biw=1366&bih=576#imgrc=eTUdAqfoV6SVe
M

41
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Driverless Car

https://www.google.co.in/search?q=googles+car&source=lnms&tbm=isch&sa=X&ved=2ahUKEwjVm9eR3cjnAhWSyDgGHRtbCSoQ_AUoAXoECA4QAw&biw=1366&bih=576#imgrc=PKVlYsSm
RZ7Z_M

42
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Industrial Robots

https://www.google.co.in/search?q=industrial+robots&tbm=isch&ved=2ahUKEwjC4LaT3cjnAhUxIrcAHVXUCukQ2-
cCegQIABAA&oq=industrial+robots&gs_l=img.3..0l10.152264.156503..156777...0.0..0.199.1705.11j6......0....1..gws-wiz-
img.......0i131j0i67.Me7iH5On5eY&ei=YzZCXoKjNrHE3LUP1airyA4&bih=576&biw=1366#imgrc=7nHoH1bOTlMf8M

43
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Cognitive ability of Human brain

Sinha, G.R. Study of assessment of cognitive ability of human brain using deep learning. Int. j. inf. tecnol. 9, 321–326
(2017). https://doi.org/10.1007/s41870-017-0025-8. https://link.springer.com/article/10.1007/s41870-017-0025-8

44
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Myanmar Sign Language Recognition

Sinha G.R., Oo P.S. (2019) Introduction to Biometrics and Special Emphasis on Myanmar Sign Language Recognition. In:
Sinha G. (eds) Advances in Biometrics. Springer, Cham. https://link.springer.com/chapter/10.1007/978-3-030-30436-2_1

45
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter
G R Sinha G R Sinha April 30, 2020 MIIT
 Brain Computer Interface (BCI)

46
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Challenges and Opportunities

 Different Soft computing Methods such as AI, Fuzzy Logic, Neural Network, Genetic Algorithm

have their own limitations and challenges

 Robustness is all time challenge. Big Data issues.

 Empathy and Societal Issues are major factors as future research in CV

 Deep Learning addresses major difficulties

Source: Google Image

47
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Concluding Remarks

 CV has tremendous capacity and scope of implementations, governed by AI

 Future is CV with challenges
 Research avenues in dealing with the challenges
 CV has to be equipped with empathy and societal values of HV
 Cognitive IoT has huge scope
 CV enables smartness: Smart Environment Monitoring, Smart City, Smart Transport System,
Smart surgery etc.

48
Computer Vision versus Human Vision @Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020
 Inspiring Equation

E= mc2

E = Excellence m= Motivation c=Commitment

Example:
c= 0.5 (half hearted), E= 1/4
&
c= 2 (doubly committed), E= 4

Thank you, any queries please!

Computer Vision versus Human Vision

View publication stats
@Hyderabad Deccan ACM Chapter G R Sinha April 30, 2020 49