w1 gemini

Question 1

What are the aims of studying Computer Vision?

Answer 1

[Implicit in later slides: Understanding the world from images

Question 2

Why is Computer Vision difficult? (Give two reasons)

Answer 2

One image → many interpretations
One object -> many images

(Exponential growth of complexity of problem with each variable)

Question 3

What are the two main approaches to tackling Computer Vision problems?

Answer 3

Computational approach
Biological approach

Question 4

Define Computer Vision according to Trucco and Verri.

Answer 4

Computing properties of the 3D world from one or more digital images

Question 5

Define Computer Vision according to Stockman and Shapiro.

Answer 5

“To make useful decisions about real physical objects and scenes based on sensed images”

Question 6

Define Computer Vision according to Forsyth and Ponce.

Answer 6

“Extracting descriptions of the world from pictures or sequences of pictures”

Question 7

Explain the relationship between Image Processing and Computer Vision.

Answer 7

Image processing manipulates images

Question 8

Explain the relationship between Computer Graphics and Computer Vision.

Answer 8

Computer graphics creates images from descriptions

Question 9

What is Pattern Recognition in the context of computer vision?

Answer 9

Recognising and classifying stimuli in images and other datasets.

Question 10

What is Photogrammetry?

Answer 10

Obtaining measurements from images.

Question 11

What is Biological Vision?

Answer 11

Understanding visual perception in humans and animals (studied in Neuroscience

Question 12

Why is vision a worthwhile subject to study?

Answer 12

Vision is the main way we experience the world.

Question 13

Give examples of applications of smile detection.

Answer 13

Cameras can be set to automatically take photos when a chosen subject laughs

Question 14

Give examples of applications of face detection/recognition.

Answer 14

Detecting faces in images to allow users to tag people.

Question 15

Give examples of biometrics applications of computer vision.

Answer 15

Iris Recognition

Question 16

Give examples of applications of people tracking.

Answer 16

Visual surveillance and crime detection (e.g.

Question 17

Give examples of applications of object tracking.

Answer 17

Instant replay and analysis in sports.

Question 18

Give examples of applications of advertising in Computer Vision.

Answer 18

Detecting the ground plane in video to introduce pictures onto it.

Question 19

Give examples of applications of content-based image retrieval.

Answer 19

Searching for images based on their content (query by image content).

Question 20

Give examples of applications of reverse image search.

Answer 20

Finding similar images.

Question 21

Give examples of applications of driver assistance systems.

Answer 21

Lane departure warning
FSD
Cruise control

Question 22

Give examples of applications of computer vision in space exploration.

Answer 22

Panorama stitching

Question 23

Give examples of applications of computer vision in medical imaging.

Answer 23

Automatic measurement and analysis of non-visual images created in MRI

Question 24

Give examples of applications of creating 3D models from images.

Answer 24

Generating a 3D virtual (CAD) model of an object or building from photos.

Question 25

Summarize why computer vision is challenging.

Answer 25

Vision is difficult because the problem is ill-posed (one image can have many interpretations) and exponentially large (one object can generate many images).

Question 26

Why is the "One image → many interpretations" problem considered ill-posed?

Answer 26

Because mapping from the 2D image back to the 3D world is not unique; many possible 3D scenes could have generated the same image.

Question 27

How do we solve the "One image → many interpretations" problem?

Answer 27

Using constraints or priors which make some interpretations more likely than others.

Question 28

Why is the "One object → many images" problem considered exponentially large?

Answer 28

Because a single object can appear with many variations in location, size, rotation etc where each extra variable multiples the number of possible interpretations

Question 29

Explain the concept of vision being an ill-posed problem.

Answer 29

There is no unique solution

Question 30

Why is vision easy for us but difficult for computers?

Answer 30

A significant portion of the cerebral cortex (~50%) is devoted to vision

Question 31

Why is it difficult to appreciate how difficult computer vision is?

Answer 31

Because our own visual processing is effortless and intuitive.

Question 32

What does it mean that all previous examples of vision systems are limited to operating in a specific (small) domain?

Answer 32

They are designed for a specific task in a specific environment and do not generalize well.

Question 33

Why are solutions to computer vision problems often not robust?

Answer 33

Because code that works well for one task might fail for different or similar tasks

Question 34

Example of non-rigid deformations affect the appearance of an object.

Answer 34

Facial expressions Movement of hand Clothing?

Question 35

Explain within-category variation in appearance.

Answer 35

Objects belonging to the same category can still have significant variations in their visual appearance.

Question 36

Explain the concept of "Discrimination despite variation" in vision.

Answer 36

The ability to recognize subtle differences between similar objects and recognize very different-looking objects as belonging to the same category.

Question 37

How do other objects in the scene affect appearance?

Answer 37

Background clutter and occlusion can make it difficult to recognize individual objects.

Question 38

What are constraints (or priors) in the context of computer vision?

Answer 38

Pre-existing knowledge or assumptions about the world that help to disambiguate interpretations of an image.

Question 39

Why are constraints/priors necessary in computer vision?

Answer 39

To solve the challenges of one image having many interpretations and one object having many images.

Question 40

Give an example of how prior knowledge can affect our perception.

Answer 40

Being able to see a camouflaged object once its content is known.

Question 41

Give an example of how prior exposure can affect our perception.

Answer 41

Change blindness

Question 42

Give an example of how context can affect our perception.

Answer 42

The ambiguous "THE CAT" example where the middle letter is interpreted correctly based on the surrounding letters.

Question 43

Explain the significance of illusions in understanding vision.

Answer 43

Illusions reveal the assumptions the visual system is making to solve the under-constrained problem of vision.

Question 44

Explain the "Forward Engineering Approach" to tackling the problem of vision.

Answer 44

Determine requirements Design system Implement Test Refine (As opposed to reverse engineering)

Question 45

What does computational (machine) vision focus on?

Answer 45

Implementing algorithms for perception.

Question 46

What does biological (human) vision focus on?

Answer 46

Modelling biological perception.