Four panels showing edge orientations: (i) diagonal edge, (ii) horizontal edge at top, (iii) vertical edge at left, (iv) axes with an arrow at angle theta Choose the correct matching:

(D) Option figure from the original question paper

(A) Option figure from the original question paper

(B) Option figure from the original question paper

(C) Option figure from the original question paper

Figure from the original question paper

(C) 1→ iii, 2→ iv, 3→ v, 4 → ii

(A) 1→ iii, 2→ iv, 3→ i, 4 → ii

(B) 1→ iii, 2→ i, 3 → ii, 4 → v

(D) 1→ iv, 2→ iii, 3→ i, 4 → ii

Figure from the original question paper

(D) None of these statements is true

Match the derivative of activation functions f(x) with their counterparts on the right column accordingly. | 1) Leaky ReLU | i) f(x)(1 - f(x)) | |---|---| | 2) Tanh | ii) 0 if x 0 | | 3) ReLU | iii) 1 - f(x)^2 | | 4) Sigmoid | iv) 0.01 if x 0 | | | v) 0 if x \geqslant 0 and 1 if x < 0 | | | vi) 0.01 if x \geqslant 0 and 1 if x < 0 |

(D) 1→ iv, 2→ iii, 3→ ii, 4 → i

(A) 1→ vi, 2→ iii, 3 →v, 4 → i

(B) 1→ i, 2→ iv, 3 → iii, 4→ ii

(C) 1→ i, 2→ v, 3 → iii, 4→ iv

Figure from the original question paper

(C) Softmax of alignment scores

(A) Sigmoid of alignment scores

(D) Linear projections of alignment scores

SegNet upsampling stage diagram: 2x2 output map (a, b, c, d) from previous decoder stage placed into a 4x4 zero map using unknown 2x2 maxpool indices from the encoder, followed by convolution with trainable decoder filters

(D) Option figure from the original question paper

(A) Option figure from the original question paper

(B) Option figure from the original question paper

(C) Option figure from the original question paper

Figure from the original question paper

(A) Option figure from the original question paper

(B) Option figure from the original question paper

(C) Option figure from the original question paper

(D) Option figure from the original question paper

Figure from the original question paper

(D) 1→ iii, 2→ i, 3 → ii, 4 →v

(A) 1→ iii, 2→ iv, 3→ ii, 4 → i

(B) 1→ i, 2→ iv, 3 → iii, 4→ ii

(C) 1→ iv, 2→ iii, 3→ ii, 4 → i

Figure from the original question paper

(A) 1→ iv, 2→ v, 3→ i, 4 → ii

(B) 1→ iii, 2→ i, 3 → v, 4→ ii

(C) 1→ iv, 2→ iii, 3→ ii, 4 → i

(D) 1→ iii, 2→ i, 3 → ii, 4→ v

Deep Learning for Computer Vision End Term PYQ May 2025: 31 Aug 2025 (Set QDB1)

Question 1

+2 marksOne correct option

Choose the correct matching:

A
B
C
D

Show answer

Correct answer

D

Question 1 on its own page

Question 2

+2 marksOne correct option

A
1→ iii, 2→ iv, 3→ i, 4 → ii
B
1→ iii, 2→ i, 3 → ii, 4 → v
C
1→ iii, 2→ iv, 3→ v, 4 → ii
D
1→ iv, 2→ iii, 3→ i, 4 → ii

Show answer

Correct answer

C
1→ iii, 2→ iv, 3→ v, 4 → ii

Question 2 on its own page

Question 3

+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

A
6→ 1→ 4→ 5 → 2
B
3→ 1→ 6→ 4 → 2
C
3→ 5→ 1→ 4 → 2
D
6→ 3→ 5→ 7 → 2

Show answer

Correct answer

B
3→ 1→ 6→ 4 → 2

Question 3 on its own page

Question 4

+2 marksOne correct option

A
Only statement P is true
B
Only statement Q is true
C
Both statements are true
D
None of these statements is true

Show answer

Correct answer

A
Only statement P is true

Question 4 on its own page

Question 5

+2 marksOne correct option

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

1) Leaky ReLU	i) $f(x)(1 - f(x))$
2) Tanh	ii) 0 if $x < 0$ and 1 if $x > 0$
3) ReLU	iii) $1 - f(x)^2$
4) Sigmoid	iv) 0.01 if $x < 0$ and 1 if $x > 0$
	v) 0 if $x \geqslant 0$ and 1 if $x < 0$
	vi) 0.01 if $x \geqslant 0$ and 1 if $x < 0$

A
1→ vi, 2→ iii, 3 →v, 4 → i
B
1→ i, 2→ iv, 3 → iii, 4→ ii
C
1→ i, 2→ v, 3 → iii, 4→ iv
D
1→ iv, 2→ iii, 3→ ii, 4 → i

Show answer

Correct answer

D
1→ iv, 2→ iii, 3→ ii, 4 → i

Question 5 on its own pageAlso asked in Quiz 1 27 Oct 2024, End Term 22 Dec 2024, Quiz 1 23 Feb 2025

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, lo-calize the key points. 2. Estimate the orientation of keypoints using the local image gra-dients.
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.

A
5, 2, 1, 3, 4
B
5, 1, 3, 2, 4
C
1, 3, 2, 4, 5
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

A
Sigmoid of alignment scores
B
ReLU of alignment scores
C
Softmax of alignment scores
D
Linear projections of alignment scores

Show answer

Correct answer

C
Softmax of alignment scores

Question 7 on its own page

Question 8

+2 marksOne correct option

Which one of the following statements is true?

A
Attention mechanisms cannot be applied to the bidirectional RNN model
B
An image captioning network cannot be trained end-to-end even though we are using 2 different modalities to train the network
C
One of the key components in the vanilla transformer are the recurrent connections that help them to deal with variable input length.
D
None of these

Show answer

Correct answer

D
None of these

Question 8 on its own pageAlso asked in Quiz 2 1 Dec 2024

Question 9

+2 marksOne correct option

A
B
C
D

Show answer

Correct answer

D

Question 9 on its own page

Question 10

+2 marksOne correct option

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

A
Validation set is the best choice for tuning hyperparameters
B
Picking hyperparameters that perform the best on the training set leads to overfitting
C
Tuning hyperparameters on the test set is an acceptable approach to find the most powerful model for generalization performance
D
Test set performance, with hyperparameters tuned on the validation set, would be a realistic measure of how the model will generalize to new data.

Show answer

Correct answer

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

Question 10 on its own pageAlso asked in End Term 31 Aug 2025

Question 11

+2 marksOne correct option

A
B
C
D

Show answer

Correct answer

A

Question 11 on its own page

Question 12

+2 marksOne correct option

A
1→ iii, 2→ iv, 3→ ii, 4 → i
B
1→ i, 2→ iv, 3 → iii, 4→ ii
C
1→ iv, 2→ iii, 3→ ii, 4 → i
D
1→ iii, 2→ i, 3 → ii, 4 →v

Show answer

Correct answer

D
1→ iii, 2→ i, 3 → ii, 4 →v

Question 12 on its own page

Question 13

+2 marksOne correct option

A
1→ iv, 2→ v, 3→ i, 4 → ii
B
1→ iii, 2→ i, 3 → v, 4→ ii
C
1→ iv, 2→ iii, 3→ ii, 4 → i
D
1→ iii, 2→ i, 3 → ii, 4→ v

Show answer

Correct answer

A
1→ iv, 2→ v, 3→ i, 4 → ii

Question 13 on its own page

Question 14

+2 marksOne correct option

What makes the Segment Anything Model (SAM) particularly advantageous for image annotation tasks compared to conventional segmentation approaches?

A
It has perfect segmentation accuracy
B
It has minimal hardware requirements for operation
C
It can generalize to segment objects it wasn’t specifically trained on
D
It automatically identifies and labels all objects in an image without user prompts

Show answer

Correct answer

C
It can generalize to segment objects it wasn’t specifically trained on

Question 14 on its own page

Question 15

+2 marksOne correct option

Statement 1: The Segment Anything Model (SAM) demonstrates zero-shot generalization capabilities across various downstream tasks.
Statement 2: SAM’s training involves a largescale, diverse mask dataset using a prompt-based segmentation approach. Which one of the following options is most appropriate?

A
Both statements are correct, and Statement 2 correctly explains Statement 1.
B
Both statements are correct, but Statement 2 does not correctly explain Statement 1.
C
Statement 1 is correct, but Statement 2 is incorrect.
D
Statement 1 is incorrect, but Statement 2 is correct.
E
Both statements are incorrect.

Show answer

Correct answer

A
Both statements are correct, and Statement 2 correctly explains Statement 1.

Question 15 on its own page

Question 16

+2 marksOne correct option

Why does DETR typically exhibit poor performance in detecting small objects compared to larger ones?

A
The CNN backbone used in DETR has a fixed receptive field that’s too large for small objects
B
The global self-attention mechanism in DETR tends to dilute the signal from small objects across all image locations
C
DETR’s loss function actively discards all detections smaller than 50x50 pixels
D
The object queries in DETR are programmed to ignore any object smaller than 10
E
DETR’s encoder-decoder architecture was specifically designed to only detect large objects

Show answer

Correct answer

B
The global self-attention mechanism in DETR tends to dilute the signal from small objects across all image locations

Question 16 on its own page

Question 17

+2 marksOne or more correct options

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

Select all that apply.

A
Generative adversarial networks (GANs) generate sharper images compared to Variational AutoEncoders (VAE)
B
GAN is an example of an implicit density estimation model
C
Fully connected layers in mapping network of Style-GAN do not change the dimension of its input
D
The generator and discriminator are always trained together in a GAN

Show answer

Correct answers

C
Fully connected layers in mapping network of Style-GAN do not change the dimension of its input
D
The generator and discriminator are always trained together in a GAN

Question 17 on its own page

Question 18

+2 marksOne or more correct options

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

Select all that apply.

A
Autoencoder are equivalent to Principal Component Analysis (PCA) provided we don’t use of non-linear activation functions
B
When using global attention on temporal data, alignment weights are learnt for encoder hidden representations for all time steps
C
Positional encoding is an important component of the transformer architecture as it conveys information about order in a given sequence
D
It is not possible to generate different captions for the same image that have similar meaning but different tone/style
E
Autoencoders can not be used for data compression as its input and output dimensions are different

Show answer

Correct answers

A
Autoencoder are equivalent to Principal Component Analysis (PCA) provided we don’t use of non-linear activation functions
B
When using global attention on temporal data, alignment weights are learnt for encoder hidden representations for all time steps
C
Positional encoding is an important component of the transformer architecture as it conveys information about order in a given sequence

Question 18 on its own page

Question 19

+2 marksOne or more correct options

Which of the following statements about CLIP are TRUE? (Select ALL that apply)

Select all that apply.

A
CLIP uses a contrastive loss function to align text and image embeddings.
B
CLIP requires image labels for training.
C
CLIP is trained to maximize cosine similarity between embeddings of matching text-image pairs.
D
CLIP model is trained with pretext tasks such as rotation prediction and inpainting to learn effective feature representations.
E
To use the CLIP model for image classification, it is imperative to finetune it on downstream datasets.

Show answer

Correct answers

A
CLIP uses a contrastive loss function to align text and image embeddings.
C
CLIP is trained to maximize cosine similarity between embeddings of matching text-image pairs.

Question 19 on its own page

Question 20

+2 marksOne or more correct options

Which one of the following statements is true:

Select all that apply.

A
Weight change criterion is a method of ‘early stopping’ that checks whether or not the error is dropping over epochs to decide whether to continue training or stop.
B
L1 norm tends to create more sparse weights than L2 norm.
C
During the training phase, for each iteration, Dropout ignores a random fraction, p, of nodes, and accounts for it in the test phase by scaling down the activations by a factor of p.
D
A single McCulloch-Pitts neuron is capable of modeling AND, OR, XOR, NOR, and NAND functions.

Show answer

Correct answers

B
L1 norm tends to create more sparse weights than L2 norm.
C
During the training phase, for each iteration, Dropout ignores a random fraction, p, of nodes, and accounts for it in the test phase by scaling down the activations by a factor of p.

Question 20 on its own pageAlso asked in End Term 31 Aug 2025

Question 21

+2 marksOne or more correct options

Which of the following are examples of a high-pass filter?

Select all that apply.

A
B
C
D

Show answer

Correct answers

A
B
C
D

Question 21 on its own page

Question 22

+2 marksOne or more correct options

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

Select all that apply.

A
B
C
D

Show answer

Correct answers

B
C

Question 22 on its own page

Question 23

+2 marksOne or more correct options

Which of the following techniques help control the exploding or vanishing gradient problem in recurrent neural networks?

Select all that apply.

A
Gradient clipping
B
Use ReLU activation funtion to control exploding gradient and sigmoid activation to control vanishing gradients
C
Use Long Short-Term Memory (LSTM)
D
Use data augmentation methods (for example, reversing the input sequence)

Show answer

Correct answers

A
Gradient clipping
C
Use Long Short-Term Memory (LSTM)
D
Use data augmentation methods (for example, reversing the input sequence)

Question 23 on its own page

Question 24

+2 marksNumerical answer

Given is a $3 \times 3$ 8-bit grayscale image:

$\begin{bmatrix} 50 & 70 & 120 \\ 90 & 30 & 80 \\ 40 & 60 & 110 \end{bmatrix}$

What is the central element after applying linear contrast stretching? Linear contrast stretching is an image enhancement technique used to improve the contrast in an image. This is achieved by linearly scaling the pixel values so that they span a desired range (0 to 255 for an 8-bit grayscale image).

Show answer

Correct answer: 0

Question 24 on its own page

Question 25

+2 marksNumerical answer

Consider the grayscale image shown below as a $5 \times 5$ matrix:

$\begin{bmatrix} 20 & 30 & 25 & 30 & 40 \\ 45 & 10 & 40 & 35 & 20 \\ 15 & 20 & 30 & 25 & 50 \\ 50 & 40 & 35 & 15 & 30 \\ 10 & 35 & 25 & 20 & 45 \end{bmatrix}$

If a median filter with a $3 \times 3$ neighborhood is applied to this image with a stride of 1 and no padding, what will be the value at position $(2, 2)$ in the filtered image? (Assume that the top-left pixel coordinate of output image is addressed as $(1, 1)$ )?

Show answer

Correct answer: 30

Question 25 on its own page

Question 26

+2 marksNumerical answer

Show answer

Correct answer: -0.21 (accepted within ±0.005)

Question 26 on its own page

Question 27

+2 marksNumerical answer

Show answer

Correct answer: 45

Question 27 on its own page

Question 28

+2 marksNumerical answer

Show answer

Correct answer: 3.75

Question 28 on its own page

Question 29

+2 marksNumerical answer

Show answer

Correct answer: 100

Question 29 on its own page

Question 30

+1 markNumerical answer

Show answer

Correct answer: -2

Question 30 on its own page

Question 31

+1 markNumerical answer

Show answer

Correct answer: 5

Question 31 on its own page

Question 32

+1 markNumerical answer

Based on the above data, answer the given subquestions.

Element 1: _______________

Show answer

Correct answer: 0.54

Question 32 on its own page

Question 33

+1 markNumerical answer

Based on the above data, answer the given subquestions.

Element 2: _______________

Show answer

Correct answer: 0.25

Question 33 on its own page

Question 34

+1 markNumerical answer

Based on the above data, answer the given subquestions.

Element 3: _______________

Show answer

Correct answer: 0.19

Question 34 on its own page

Question 35

+1 markNumerical answer

Based on the above data, answer the given subquestions.

Element 4: _______________

Show answer

Correct answer: 0.58

Question 35 on its own page

Question 36

+1 markNumerical answer

A 4-dimensional input vector $x = [5, 3, -1, 2]$ is passed to a hidden layer with a single neuron and an activation function $a(.)$ to obtain $z$ . Assume that the corresponding weights (from input to hidden layer) are $[0.4, 0.6, 0.8, -0.2]$ and bias is $0.4$ . If the activation function $a(.)$ is Sigmoid, Linear, Indicator function, Softplus, ReLU and Leaky- ReLU, what are the corresponding values of $z$ ? (Note that Indicator function returns 1 for a positive input, and 0 otherwise. Softplus is defined as: $\ln(1 + \exp(z))$ .

Based on the above data, answer the given subquestions.

Sigmoid _______________

Show answer

Correct answer: 0.95

Question 36 on its own page

Question 37

+1 markNumerical answer

A 4-dimensional input vector $x = [5, 3, -1, 2]$ is passed to a hidden layer with a single neuron and an activation function $a(.)$ to obtain $z$ . Assume that the corresponding weights (from input to hidden layer) are $[0.4, 0.6, 0.8, -0.2]$ and bias is $0.4$ . If the activation function $a(.)$ is Sigmoid, Linear, Indicator function, Softplus, ReLU and Leaky- ReLU, what are the corresponding values of $z$ ? (Note that Indicator function returns 1 for a positive input, and 0 otherwise. Softplus is defined as: $\ln(1 + \exp(z))$ .

Based on the above data, answer the given subquestions.

Linear _______________

Show answer

Correct answer: 3

Question 37 on its own page

Question 38

+1 markNumerical answer

A 4-dimensional input vector $x = [5, 3, -1, 2]$ is passed to a hidden layer with a single neuron and an activation function $a(.)$ to obtain $z$ . Assume that the corresponding weights (from input to hidden layer) are $[0.4, 0.6, 0.8, -0.2]$ and bias is $0.4$ . If the activation function $a(.)$ is Sigmoid, Linear, Indicator function, Softplus, ReLU and Leaky- ReLU, what are the corresponding values of $z$ ? (Note that Indicator function returns 1 for a positive input, and 0 otherwise. Softplus is defined as: $\ln(1 + \exp(z))$ .

Based on the above data, answer the given subquestions.

Indicator Function _______________

Show answer

Correct answer: 1

Question 38 on its own page

Question 39

+1 markNumerical answer

A 4-dimensional input vector $x = [5, 3, -1, 2]$ is passed to a hidden layer with a single neuron and an activation function $a(.)$ to obtain $z$ . Assume that the corresponding weights (from input to hidden layer) are $[0.4, 0.6, 0.8, -0.2]$ and bias is $0.4$ . If the activation function $a(.)$ is Sigmoid, Linear, Indicator function, Softplus, ReLU and Leaky- ReLU, what are the corresponding values of $z$ ? (Note that Indicator function returns 1 for a positive input, and 0 otherwise. Softplus is defined as: $\ln(1 + \exp(z))$ .

Based on the above data, answer the given subquestions.

Softplus _______________

Show answer

Correct answer: 3.05

Question 39 on its own page

Question 40

+1 markNumerical answer

A 4-dimensional input vector $x = [5, 3, -1, 2]$ is passed to a hidden layer with a single neuron and an activation function $a(.)$ to obtain $z$ . Assume that the corresponding weights (from input to hidden layer) are $[0.4, 0.6, 0.8, -0.2]$ and bias is $0.4$ . If the activation function $a(.)$ is Sigmoid, Linear, Indicator function, Softplus, ReLU and Leaky- ReLU, what are the corresponding values of $z$ ? (Note that Indicator function returns 1 for a positive input, and 0 otherwise. Softplus is defined as: $\ln(1 + \exp(z))$ .

Based on the above data, answer the given subquestions.

ReLu _______________

Show answer

Correct answer: 3

Question 40 on its own page

Question 41

+1 markNumerical answer

A 4-dimensional input vector $x = [5, 3, -1, 2]$ is passed to a hidden layer with a single neuron and an activation function $a(.)$ to obtain $z$ . Assume that the corresponding weights (from input to hidden layer) are $[0.4, 0.6, 0.8, -0.2]$ and bias is $0.4$ . If the activation function $a(.)$ is Sigmoid, Linear, Indicator function, Softplus, ReLU and Leaky- ReLU, what are the corresponding values of $z$ ? (Note that Indicator function returns 1 for a positive input, and 0 otherwise. Softplus is defined as: $\ln(1 + \exp(z))$ .

Based on the above data, answer the given subquestions.

Leaky-ReLu _______________

Show answer

Correct answer: 3

Question 41 on its own page

Question 42

+0.55 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 1

Question 42 on its own page

Question 43

+0.55 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 5

Question 43 on its own page

Question 44

+0.55 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 11

Question 44 on its own page

Question 45

+0.55 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 7

Question 45 on its own page

Question 46

+0.55 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 8

Question 46 on its own page

Question 47

+0.55 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 4

Question 47 on its own page

Question 48

+0.54 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 10

Question 48 on its own page

Question 49

+0.54 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 9

Question 49 on its own page

Question 50

+0.54 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 3

Question 50 on its own page

Question 51

+0.54 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 2

Question 51 on its own page

Question 52

+0.54 marksNumerical answer

Consider the BLIP model architecture with an input (image-text pair) as shown below: For each of the given entities, enter the appropriate number from the figure that corresponds to it:

Based on the above data, answer the given subquestions.

Show answer

Correct answer: 6

Question 52 on its own page