uiz Space

May 2025 term · Deep Learning for Computer Vision · BSDA5006

Deep Learning for Computer Vision Quiz 1: 13 July 2025 (May 2025 term)

The IIT Madras BS Deep Learning for Computer Vision (Deep Learning for Computer Vision) Quiz 1 paper sat on 13 Jul 2025, in the May 2025 term: 34 questions for 79 marks in 120 minutes. Every question is below with its answer. Take it as a timed mock test to be marked, or read it through first.

Questions
34
Marks
79
Duration
120 min
MCQ
14
MSQ
5
Numerical
11
Written
4
Take as mock testPractise with answers

Updated

Official paper: IIT M DEGREE AN EXAM QDB2 13 July 2025 · No negative marking.

Question 1

+2 marksOne correct option
  1. A

    104.94

  2. B

    100.25

  3. C

    110.56

  4. D

    98.78

Show answer

Correct answer

  • A

    104.94

Question 1 on its own page

Question 2

+2 marksOne correct option

Which of the following statements is false?

  1. A

    Histogram equalization is a global operation.

  2. B

    Gaussian filtering is a local operation.

  3. C

    Convolution in the spatial domain corresponds to multiplication in the frequency domain.

  4. D

    Median filtering is a global operation.

Show answer

Correct answer

  • D

    Median filtering is a global operation.

Question 2 on its own page

Question 3

+2 marksOne correct option

Which of the following statements is True?

  1. A

    Gaussian filter is a separable filter because it is linear.

  2. B

    Median filter is a non-separable filter.

  3. C

    Gaussian filter is a high-pass filter.

  4. D

    Mean filter is a non-separable filter because it is non-linear.

Show answer

Correct answer

  • A

    Gaussian filter is a separable filter because it is linear.

Question 3 on its own page

Question 4

+2 marksOne correct option
  1. A
  2. B
  3. C
  4. D
Show answer

Correct answer

  • A
Question 4 on its own page

Question 5

+2 marksOne correct option
  1. A

    0.9820, 4.00, 1, 4.02, 4.00, 4.00

  2. B

    0.8808, 3.50, 1, 3.54, 3.50, 3.50

  3. C

    0.7311, 2.00, 1, 2.13, 2.00, 2.00

  4. D

    0.5000, 0.00, 0, 0.69, 0.00, 0.00

Show answer

Correct answer

  • A

    0.9820, 4.00, 1, 4.02, 4.00, 4.00

Question 5 on its own page

Question 6

+2 marksOne correct option

Identify the correct sequence of steps in a Canny edge detection pipeline. Steps are listed below:

  1. Compute gradient magnitude and direction
  2. Connect individual components
  3. Smoothen the image
  4. Threshold into strong, weak, or no edge
  5. Gaussian Filter and Hysteresis
  6. Non-maximum suppression
  7. Apply derivative to get edges
  1. A
  2. B
  3. C
  4. D
Show answer

Correct answer

  • B
Question 6 on its own page

Question 7

+2 marksOne correct option
  1. A
  2. B
  3. C
  4. D
Show answer

Correct answer

  • B
Question 7 on its own page

Question 8

+2 marksOne correct option

During the double thresholding step in Canny edge detection, the high threshold is set at 100 and the low threshold is 40. A pixel gradient magnitude of 65 will be classified as:

  1. A

    Strong edge

  2. B

    Weak edge

  3. C

    Non-edge

  4. D

    Ambiguous edge

Show answer

Correct answer

  • B

    Weak edge

Question 8 on its own page

Question 9

+2 marksOne correct option

Match the derivative of activation functions f(x)f(x) with their counterparts on the right column accordingly.

1) Leaky ReLUi) f(x)(1−f(x))f(x)(1 - f(x))
2) Tanhii) 0 if x<0x < 0 and 1 if x>0x > 0
3) ReLUiii) 1−f(x)21 - f(x)^2
4) Sigmoidiv) 0.01 if x<0x < 0 and 1 if x>0x > 0
v) 0 if x⩾0x \geqslant 0 and 1 if x<0x < 0
vi) 0.01 if x⩾0x \geqslant 0 and 1 if x<0x < 0
  1. A
  2. B
  3. C
  4. D
Show answer

Correct answer

  • D
Question 9 on its own page

Question 10

+2 marksOne correct option
  1. A
  2. B
  3. C
  4. D
Show answer

Correct answer

  • C
Question 10 on its own page

Question 11

+2 marksOne correct option

Which of the following statements are True?

  1. A

    Vanilla Gradient Descent converges faster than Momentum-based GD.

  2. B

    Momentum based GD oscillates around minima before convergence.

  3. C

    Noise in Stochastic Gradient descent weight updates – can lead to faster convergence.

  4. D

    None of these

Show answer

Correct answer

  • B

    Momentum based GD oscillates around minima before convergence.

Question 11 on its own pageAlso asked in Quiz 1 23 Feb 2025

Question 12

+2 marksOne correct option
  1. A

    0.16

  2. B

    0.08

  3. C

    0.20

  4. D

    0.12

Show answer

Correct answer

  • A

    0.16

Question 12 on its own page

Question 13

+2 marksOne correct option
  1. A
  2. B
  3. C
  4. D
Show answer

Correct answer

  • C
Question 13 on its own page

Question 14

+2 marksOne correct option
  1. A
  2. B
  3. C
  4. D
Show answer

Correct answer

  • A
Question 14 on its own page

Question 15

+3 marksOne or more correct options

Which of the following statements are true?(Select all that apply)

Select all that apply.

  1. A
  2. B
  3. C
  4. D
Show answer

Correct answers

  • A
  • D
Question 15 on its own page

Question 16

+3 marksOne or more correct options

Which of the following statements are true ?(Select all that apply)

Select all that apply.

  1. A
  2. B
  3. C
  4. D
Show answer

Correct answers

  • A
  • B
Question 16 on its own page

Question 17

+3 marksOne or more correct options

Consider the following statements. Which of the following statements are true?

Select all that apply.

  1. A
  2. B
  3. C
  4. D
  5. E
  6. F
Show answer

Correct answers

  • A
  • C
  • E
  • F
Question 17 on its own page

Question 18

+3 marksOne or more correct options

Select all that apply.

  1. A
  2. B
  3. C
  4. D
Show answer

Correct answers

  • A
  • D
Question 18 on its own page

Question 19

+3 marksOne or more correct options

Which of the following statements are false?

Select all that apply.

  1. A

    Momentum in optimization can cause oscillations around minima when encountering flat regions or saddle points due to the dominance of the momentum term over the small gradient.

  2. B

    Stochastic Gradient Descent (SGD) with its inherent noise can be beneficial in escaping local minima and saddle points, provided that there is enough gradient information in the neighborhood.

  3. C

    Adagrad is an optimization algorithm introduced to overcome the diminishing learning rate problem in techniques like RMSProp.

  4. D

    ADAM is introduced to solve problems in RMSProp by combining RMSProp and Adagrad techniques.

Show answer

Correct answers

  • C

    Adagrad is an optimization algorithm introduced to overcome the diminishing learning rate problem in techniques like RMSProp.

  • D

    ADAM is introduced to solve problems in RMSProp by combining RMSProp and Adagrad techniques.

Question 19 on its own pageAlso asked in Quiz 1 27 Oct 2024, Quiz 1 23 Feb 2025

Question 20

+3 marksNumerical answer

Consider the grayscale image shown below:

[12182225302015172123251920242827222318203024211925]\begin{bmatrix} 12 & 18 & 22 & 25 & 30 \\ 20 & 15 & 17 & 21 & 23 \\ 25 & 19 & 20 & 24 & 28 \\ 27 & 22 & 23 & 18 & 20 \\ 30 & 24 & 21 & 19 & 25 \end{bmatrix}

If a median filter with a 3×33 \times 3 neighborhood is applied to this image with a stride of 1 and no padding, __________ will be the value at position (3,3)(3, 3) in the filtered image? (Assume that the top-left pixel coordinate of the output image is addressed as (1,1)(1, 1).)

Show answer

Correct answer: 20

Question 20 on its own page

Question 21

+3 marksNumerical answer
Show answer

Correct answer: 4

Question 21 on its own page

Question 22

+3 marksNumerical answer

Given a 2×22 \times 2 second moment matrix of edge pixels:

M=[1221]M = \begin{bmatrix} 1 & 2 \\ 2 & 1 \end{bmatrix}

obtained during the Canny edge detection process, compute the measure of straightness defined as:

Straightness=λmin⁡−λmax⁡λmax⁡+λmin⁡\text{Straightness} = \frac{\lambda_{\min} - \lambda_{\max}}{\lambda_{\max} + \lambda_{\min}}

where λmax⁡\lambda_{\max} and λmin⁡\lambda_{\min} are the maximum and minimum eigenvalues of MM respectively.

The value of the straightness is __________

Show answer

Correct answer: -2

Question 22 on its own page

Question 23

+3 marksNumerical answer
Show answer

Correct answer: 28

Question 23 on its own page

Question 24

+3 marksNumerical answer
Show answer

Correct answer: 8

Question 24 on its own page

Question 25

+3 marksNumerical answer
Show answer

Correct answer: 86400

Question 25 on its own page

Question 26

+3 marksNumerical answer
Show answer

Correct answer: -0.235 (accepted within ±0.025)

Question 26 on its own page

Question 27

+3 marksNumerical answer

Consider the following numpy array

arr=[5678]\text{arr} = \begin{bmatrix} 5 & 6 \\ 7 & 8 \end{bmatrix}

Now consider the following operations

copied_arr=arr.copy()\text{copied\_arr} = \text{arr.copy()}

copied_arr[0, 0]=100\text{copied\_arr[0, 0]} = 100

Write the value of arr[0, 0]

Show answer

Correct answer: 5

Question 27 on its own page

Question 28

+3 marksNumerical answer

Consider the following numpy array

arr=[1234]\text{arr} = \begin{bmatrix} 1 & 2 \\ 3 & 4 \end{bmatrix}

Now consider the following operations

viewed_arr=arr.view()\text{viewed\_arr} = \text{arr.view()}

viewed_arr[0, 1]=99\text{viewed\_arr[0, 1]} = 99

Write the value of arr[0, 1]

Show answer

Correct answer: 99

Question 28 on its own page

Question 29

+1.5 marksWritten answer

The gradient of an image I points in the direction of the most rapid change in intensity. Taking image derivatives accentuates (a)_________________ frequencies and hence amplifies noise, since the proportion of noise to signal is larger at (b) _________________frequencies. The common solution is to smooth the image prior to computing gradients.
Based on the above data, answer the given subquestions.

Enter the correct answer for Blank (a):
NOTE: Enter the exact answer without any extra space in the beginning or at the end.

Show answer

Correct answer: high or higher or large or largest

Question 29 on its own page

Question 30

+1.5 marksWritten answer

The gradient of an image I points in the direction of the most rapid change in intensity. Taking image derivatives accentuates (a)_________________ frequencies and hence amplifies noise, since the proportion of noise to signal is larger at (b) _________________frequencies. The common solution is to smooth the image prior to computing gradients.
Based on the above data, answer the given subquestions.

Enter the correct answer for Blank (b):
NOTE: Enter the exact answer without any extra space in the beginning or at the end.

Show answer

Correct answer: high or higher or large or largest

Question 30 on its own page

Question 31

+1.5 marksWritten answer

In Canny edge detection, a large value of the Gaussian kernel spread σ leads to (a)________________ edges, while a small value of σ leads to (b)_________________ edges.
Based on the above data, answer the given subquestions.

Enter the correct answer for Blank (a) :
NOTE: Enter the exact answer without any extra space in the beginning or at the end.

Show answer

Correct answer: thick or coarse or large-scale

Question 31 on its own page

Question 32

+1.5 marksWritten answer

In Canny edge detection, a large value of the Gaussian kernel spread σ leads to (a)________________ edges, while a small value of σ leads to (b)_________________ edges.
Based on the above data, answer the given subquestions.

Enter the correct answer for Blank (b) :
NOTE: Enter the exact answer without any extra space in the beginning or at the end.

Show answer

Correct answer: fine or thin

Question 32 on its own page

Question 33

+1.5 marksNumerical answer

You have a visual codebook with 1500 visual words. From a test image, you extract 1200 SIFT descriptors, each of 128 dimensions. After assigning each descriptor to its nearest visual word, you compute the Bag-of-Words histogram by counting descriptor assignments.
Based on the above data, answer the given subquestions.

What is the dimensionality of the Bag-of-Words (BoW) feature vector for this
image?________________

Show answer

Correct answer: 1500

Question 33 on its own page

Question 34

+1.5 marksNumerical answer

You have a visual codebook with 1500 visual words. From a test image, you extract 1200 SIFT descriptors, each of 128 dimensions. After assigning each descriptor to its nearest visual word, you compute the Bag-of-Words histogram by counting descriptor assignments.
Based on the above data, answer the given subquestions.

If instead, you concatenate all the 1200 descriptors into a single vector (without using BoW), what will be the dimensionality of that vector?__________________

Show answer

Correct answer: 153600

Question 34 on its own page