Quiz Space

May 2024 term · Deep Learning for Computer Vision · BSDA5006

Deep Learning for Computer Vision End Term: 1 September 2024 (May 2024 term)

The IIT Madras BS Deep Learning for Computer Vision (Deep Learning for Computer Vision) End Term paper sat on 1 Sept 2024, in the May 2024 term: 34 questions for 50 marks in 180 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
50
Duration
180 min
MCQ
15
Numerical
7
Written
9
MSQ
3
Take as mock testPractise with answers

Updated

Official paper: IIT M DEGREE AN EXAM QDB3 01 Sep 2024 · No negative marking.

Question 1

+2 marksOne correct option

What is the role of the Gaussian blur in the Canny edge detection algorithm, and how does the choice of the standard deviation parameter affect the performance of edge detection? Which option do you think best describes the role of the Gaussian blur and its relationship with the standard deviation parameter in the Canny edge detection algorithm?

  1. A

    The Gaussian blur helps to reduce noise in the image and smooth out pixel intensity variations.The standard deviation parameter controls the amount of blurring applied, with higher values resulting in more aggressive smoothing, which may lead to loss of edge detail.

  2. B

    The Gaussian blur enhances edge contrast in the image and amplifies pixel intensity gradients. The standard deviation parameter determines the width of the blur kernel, with larger values resulting in sharper edges and finer details preserved.

  3. C

    The Gaussian blur serves to highlight high-frequency components in the image and accentuate edge boundaries. The standard deviation parameter dictates the level of detail preserved, with smaller values preserving finer details but potentially amplifying noise.

  4. D

    The Gaussian blur functions to dilate edge regions in the image and expand the edge boundaries.The standard deviation parameter controls the extent of dilation, with higher values resulting in broader edge regions but also increasing the risk of false edge detections.

Show answer

Correct answer

  • A

    The Gaussian blur helps to reduce noise in the image and smooth out pixel intensity variations.The standard deviation parameter controls the amount of blurring applied, with higher values resulting in more aggressive smoothing, which may lead to loss of edge detail.

Question 1 on its own pageAlso asked in End Term 28 Apr 2024

Question 2

+2 marksOne correct option

Consider the following orientation of edges s:

Choose the correct matching:

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

Correct answer

  • D
Question 2 on its own page

Question 3

+2 marksOne correct option

Match the following:

  1. A

    1→ iii, 2→ iv, 3 →i, 4 → ii

  2. B

    1→ iii, 2→ i, 3 → ii, 4 → v

  3. C

    1→ iii, 2→ iv, 3 →v, 4 → ii

  4. D

    1→ iv, 2→ iii, 3 → i, 4→ ii

Show answer

Correct answer

  • C

    1→ iii, 2→ iv, 3 →v, 4 → ii

Question 3 on its own page

Question 4

+2 marksOne correct option

Consider the following statements and choose the correct option:
(i) Grad-CAM overcame the problems of previous works to work better in images with occlusion and multiple instances of objects.
(ii) Grad-CAM ++ uses a weighted average of pixels to solve the problem of activation maps with lesser spatial footprint getting suppressed.

  1. A

    Only statement (i) is True

  2. B

    Only statement (ii) is True

  3. C

    Both the statements are True

  4. D

    None of the statements is True

Show answer

Correct answer

  • D

    None of the statements is True

Question 4 on its own page

Question 5

+2 marksOne correct option

Which one of the following statements regarding hyperparameter tuning is True?

  1. A

    Test set performance, with hyperparameters tuned on the validation set, would be a realistic measure of how the model will generalize to new data.

  2. B

    Train set is the best choice for tuning hyperparameters

  3. C

    Picking hyperparameters that perform the best on the validation set leads to overfitting

  4. D

    Tuning hyperparameters on the test set is an acceptable approach to find the most powerful model for generalization performance

Show answer

Correct answer

  • A

    Test set performance, with hyperparameters tuned on the validation set, would be a realistic measure of how the model will generalize to new data.

Question 5 on its own page

Question 6

+2 marksOne correct option

Which of the following is the correct sequence of steps of the SIFT algorithm?
1. Using the Taylor series expansion of the scale-space function, localize the key points. 2. Estimate the orientation of keypoints using the local image gradients.
3. Find the edge points using the eigenvalues of the Hessian and reject these points.
4. Find a raw SIFT descriptor vector using the gradient orientations of the neighborhood around the key points.
5. Detect extreme points in scale space using an approximation of the Laplacian of Gaussian.

  1. A

    5, 2, 1, 3, 4

  2. B

    5, 1, 3, 2, 4

  3. C

    1, 3, 2, 4, 5

  4. D

    1, 5, 3, 2, 4

Show answer

Correct answer

  • B

    5, 1, 3, 2, 4

Question 6 on its own page

Question 7

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

Correct answer

  • A
Question 7 on its own page

Question 8

+2 marksOne correct option

Which one of the following distance measures is used to compare two images in dense registration through optical flow?

  1. A

    Weighted squared difference

  2. B

    Weighted absolute difference

  3. C

    KL divergence

  4. D

    Cosine similarity

Show answer

Correct answer

  • A

    Weighted squared difference

Question 8 on its own page

Question 9

+2 marksOne correct option

Which one of the following statements is false?

  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

What can we do with a generative model?
Match the following (select one among the options provided below):

1) Discriminative modeli) Assign labels to data; Feature learning (supervised)
2) Generative modelii) Detect outliers; Feature learning (unsupervised); Sample to generate new data
3) Conditional generative modeliii) Assign labels, while rejecting outliers; Generate new data conditioned on input labels
  1. A

    1→iii,2→i, 3→ii

  2. B

    1→i,2→iii, 3→ii

  3. C

    1→ii,2→iii, 3→i

  4. D

    1→i,2→ii, 3→iii

Show answer

Correct answer

  • D

    1→i,2→ii, 3→iii

Question 10 on its own page

Question 11

+2 marksOne correct option
  1. A

    21.3%

  2. B

    12.8%

  3. C

    11.5%

  4. D

    8.7%

Show answer

Correct answer

  • D

    8.7%

Question 11 on its own page

Question 12

+2 marksNumerical answer
Show answer

Correct answer: 4

Question 12 on its own page

Question 13

+2 marksNumerical answer
Show answer

Correct answer: 18

Question 13 on its own page

Question 14

+2 marksNumerical answer
Show answer

Correct answer: 14

Question 14 on its own page

Question 15

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 1
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Decoder Masked Multi-Head Attention

Question 15 on its own page

Question 16

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 2
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Encoder Feed Forward Layer

Question 16 on its own page

Question 17

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 3
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Output Embedding

Question 17 on its own page

Question 18

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 4
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Softmax

Question 18 on its own page

Question 19

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 5
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Linear projection

Question 19 on its own page

Question 20

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 6
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Input Embedding

Question 20 on its own page

Question 21

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 7
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Decoder Feed Forward Layer

Question 21 on its own page

Question 22

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 8
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Encoder Multi-Head Attention

Question 22 on its own page

Question 23

+1 markWritten answer

Consider a vanilla transformer network at time step ’t’ shown below:

For each of the entities above, enter the appropriate Model component from the figure that corresponds to it:
Based on the above data, answer the given subquestions.

Enter the answer for Model component 9
NOTE: Enter the exact answer without any space in the beginning or at the end.

Show answer

Correct answer: Decoder Multi-Head Attention

Question 23 on its own page

Question 24

+0.67 marksNumerical answer

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 0.301

Question 24 on its own page

Question 25

+0.67 marksNumerical answer

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 0.3

Question 25 on its own page

Question 26

+0.66 marksNumerical answer

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 0.25

Question 26 on its own page

Question 27

+2 marksOne or more correct options

Which one of the following statements is true?

Select all that apply.

  1. A

    Attention mechanisms cannot be applied to the bidirectional RNN model

  2. B

    An image captioning network cannot be trained end-to-end even though we are using 2 different modalities to train the network

  3. C

    One of the key components in the vanilla transformer are the recurrent connections that help them to deal with variable input length.

  4. D

    None of these

Show answer

Correct answer

  • D

    None of these

Question 27 on its own page

Question 28

+2 marksOne or more correct options

Which of the following statements are true? (Select all possible correct options)

Select all that apply.

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

Correct answers

  • A
  • C
Question 28 on its own page

Question 29

+2 marksOne or more correct options

In adversarial robustness literature, which of the following are white box attacks?

Select all that apply.

  1. A

    Carlini and Wagner (C&W) attack

  2. B

    Projected Gradient Descent (PGD)

  3. C

    Random Input Transformation

  4. D

    Zeroth-Order Optimization (ZOO)

  5. E

    Opt-Attack

Show answer

Correct answers

  • A

    Carlini and Wagner (C&W) attack

  • B

    Projected Gradient Descent (PGD)

Question 29 on its own page

Question 30

+1 markNumerical answer

Consider a Faster R-CNN object detection framework, which is being trained with a minibatch size of 2. The RPN produces 512 proposals, out of which 5 and 8 of them are foreground proposals, corresponding to each image in the mini-batch. For a single mini-batch update, the bounding box regression loss will be computed ______________________ times.

Show answer

Correct answer: 13

Question 30 on its own page

Question 31

+1 markOne correct option

Match the following computer vision tasks to situations:

  1. A

    1→ iv, 2→ iii, 3→ ii, 4 → i

  2. B

    1→ ii, 2→ iii, 3 → i, 4 →iv

  3. C

    1→ iv, 2→ iii, 3→ i, 4 → ii

  4. D

    1→ ii, 2→ iv, 3 → iii, 4→ i

Show answer

Correct answer

  • C

    1→ iv, 2→ iii, 3→ i, 4 → ii

Question 31 on its own page

Question 32

+1 markOne correct option

Which of the following tasks are the Vision Language tasks?

  1. A

    Image retrieval from text

  2. B

    Object detection from image and text

  3. C

    Visual Question Answering

  4. D

    All of these

Show answer

Correct answer

  • D

    All of these

Question 32 on its own page

Question 33

+1 markOne correct option

Generative Adversarial Networks (GANs) are widely used to generate new images. Many GAN- based architectures have been developed over the years to improve the quality of
generation.Progressive GAN, a popular architecture among many others, uses a technique called progressive growing. Which one of the options below is the purpose for progressive growing?

  1. A

    To grow the range of loss function

  2. B

    To gradually train the generator by increasing the resolution of images being generated in iterations

  3. C

    To iteratively train the generator to create 3D images from 2D ones

  4. D

    To shrink the size of the generator model so generator become faster

Show answer

Correct answer

  • B

    To gradually train the generator by increasing the resolution of images being generated in iterations

Question 33 on its own page

Question 34

+1 markOne correct option

Which of the following is not a self-supervised learning method?

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

Correct answer

  • B
Question 34 on its own page