Indian Institute of Technology Madras

BS Degree Programme · Examination & Study Paper

Deep Learning for Computer Vision

All 2025 papers

Official Paper

Examination Instructions & Guidelines

Solve on notebook/paper first before checking solutions.
MCQ: Exactly 1 option is correct. Negative marking applies.
MSQ: One or more options correct. Select all to score.
NAT: Numerical response. No negative marking.
Scientific calculator allowed for all calculations.
Official answer key is sealed at the end of this sheet.

Quiz 1 · 26 Oct 2025 (AN)Quiz 1

Q1MCQ2 Marks·Choose one option
Asked 3 times Key ↓
1. A130.00
2. B127.50
3. C125.00
4. D120.00
Q2MCQ2 Marks·Choose one option
Key ↓
Which of the following statements is false?
1. ABilateral filtering is a non-linear, edge-preserving local operation.
2. BMedian filtering is a linear filtering operation.
3. CHistogram matching uses CDFs to map intensities between images.
4. DUnsharp masking implements a high-boost (high-pass) effect.
Q3MCQ2 Marks·Choose one option
Key ↓
Which of the following statements is True?
1. A
2. B
3. C
4. D
Q4MCQ2 Marks·Choose one option
Key ↓
Which filter leaves the center pixel unchanged for any image under convolution?
1. A
2. B
3. C
4. D
Q5MCQ2 Marks·Choose one option
Asked 13 times Key ↓
1. A0.9829, 4.05, 1, 4.0673, 4.05, 4.05
2. B0.8808, 3.50, 1, 3.54, 3.50, 3.50
3. C0.7311, 2.00, 1, 2.13, 2.00, 2.00
4. D0.5000, 0.00, 0, 0.69, 0.00, 0.00
Q6MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Choose the correct sequence for Canny edge detection: 1. Smooth with Gaussian 2. Compute gradient magnitude & direction 3. Non-maximum suppression 4. Double thresholding (strong/weak) 5. Edge tracking by hysteresis
1. A2 → 1 → 3 → 4 → 5
2. B1 → 2 → 3 → 4 → 5
3. C1 → 3 → 2 → 4 → 5
4. D3 → 1 → 2 → 5 → 4
Q7MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Correct order of SIFT stages is:
1. AOrientation→ Descriptor → Localization → Extrema Detection
2. BExtrema Detection → Localization→ Orientation→ Descriptor
3. CExtrema Detection → Descriptor → Localization → Orientation
4. DNone of these
Q8MCQ2 Marks·Choose one option
Key ↓
In double thresholding, T_(high) = 120, T_(low) = 50. A pixel with gradient magnitude 45 is classified as:
1. AStrong edge
2. BWeak edge
3. CNon-edge
4. DAmbiguous
Q9MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A1→ii, 2→i, 3→iii, 4→iv
2. B1→i, 2→ii, 3→iii, 4→iv
3. C1→i, 2→iv, 3→ii, 4→iii
4. D1→iv, 2→ii, 3→i, 4→iii
Q10MCQ2 Marks·Choose one option
Key ↓
An RGB image goes through Conv layer: 64 kernels of size 7 x 7, stride 2, then MaxPool 3 x 3, stride 2. The receptive field of a pooling-layer neuron is:
1. A7 x 7
2. B9 x 9
3. C11 x 11
4. D13 x 13
Q11MCQ2 Marks·Choose one option
Key ↓
Which is True?
1. ANesterov momentum looks ahead and often converges faster than classical momentum.
2. BA large fixed learning rate always speeds convergence.
3. CAdam uses only a first-moment estimate of gradients.
4. DNone of these
Q12MCQ2 Marks·Choose one option
Key ↓
1. A0.066
2. B0.071
3. C0.083
4. D0.059
Q13MCQ2 Marks·Choose one option
Asked 3 times Key ↓
1. A1→iii, 2→i, 3→v, 4→iv
2. B1→i, 2→iii, 3→v, 4→ii
3. C1→iii, 2→iv, 3→v, 4→ii
4. D1→iii, 2→iv, 3→ii, 4→v
Q14MCQ2 Marks·Choose one option
Key ↓
1. A1→iii, 2→i, 3→ii
2. B1→ii, 2→iii, 3→i
3. C1→iii, 2→ii, 3→i
4. DNone of these
Q15MSQ3 Marks·Choose all correct options
Key ↓
Which statements are true? (Select all that apply)
1. A
2. B
3. C
4. D
Q16MSQ3 Marks·Choose all correct options
Key ↓
Which statements are true? (Select all that apply)
1. A
2. B
3. C
4. D
Q17MSQ3 Marks·Choose all correct options
Key ↓
About convolution and Fourier analysis, which are true?
1. A
2. B
3. C
4. D
Q18MSQ3 Marks·Choose all correct options
Asked 3 times Key ↓
Which of the following 3 x 3 kernels are separable?
1. A
2. B
3. C
4. D
Q19MSQ3 Marks·Choose all correct options
Key ↓
Which statements are false?
1. AMomentum maintains an EMA-like velocity of gradients/updates.
2. BRMSProp normalizes step sizes by a running RMS of gradients.
3. CIn the Adam optimizer, the hyper-parameters β_1 and β_2 control the exponential decay rates for the moving averages of the gradient and its squared vale. Adam with reduces to SGD with momentum. β_1 = 0 and β_2 = 0 reduces to SGD with momentum.
4. DCosine-annealed warm restarts can help escape sharp minima.
Q20NAT3 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q21NAT3 Marks·Write the number
Key ↓
Your answer:
Q22NAT3 Marks·Write the number
Key ↓
Your answer:
Q23NAT3 Marks·Write the number
Key ↓
Your answer:
Q24NAT3 Marks·Write the number
Asked 3 times Key ↓
Two successive convolutions each with: kernel 7 x 7, padding 3, stride 2. Input is 100 x 100. The final feature-map side length F is _________________.
Your answer:
Q25NAT3 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q26NAT3 Marks·Write the number
Asked 3 times Key ↓
NOTE: Enter your answer in three decimal places.
Your answer:
Q27NAT3 Marks·Write the number
Key ↓
Your answer:
Q28NAT3 Marks·Write the number
Key ↓
Your answer:
Read the passage below, then answer Q29–Q30
In Canny, increasing the high threshold tends to (a) _______________ false positives and (b) ____________ false negatives.
Based on the above data, answer the given subquestions.
Q29Short answer1.5 Marks·Write the answer
Key ↓
Enter the correct answer for Blank (a) ________________ NOTE: Enter the exact answer without any extra space in the beginning or at the end.
Your answer:
Q30Short answer1.5 Marks·Write the answer
Key ↓
Enter the correct answer for Blank (b) ________________ NOTE: Enter the exact answer without any extra space in the beginning or at the end.
Your answer:
Read the passage below, then answer Q31–Q32
During non-maximum suppression, comparisons are made along the gradient (a) _____________ to keep only local (b) _______________.
Based on the above data, answer the given subquestions.
Q31Short answer1.5 Marks·Write the answer
Key ↓
Enter the correct answer for Blank (a) ________________ NOTE: Enter the exact answer without any extra space in the beginning or at the end.
Your answer:
Q32Short answer1.5 Marks·Write the answer
Key ↓
Enter the correct answer for Blank (b) ________________ NOTE: Enter the exact answer without any extra space in the beginning or at the end.
Your answer:
Read the passage below, then answer Q33–Q34
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.
Q33Short answer1.5 Marks·Write the answer
Key ↓
Enter the correct answer for Blank (a) ________________ NOTE: Enter the exact answer without any extra space in the beginning or at the end.
Your answer:
Q34Short answer1.5 Marks·Write the answer
Key ↓
Enter the correct answer for Blank (b) ________________ NOTE: Enter the exact answer without any extra space in the beginning or at the end.
Your answer:

End Term · 31 Aug 2025 (FN)End Term

Q35MCQ2 Marks·Choose one option
Key ↓
Choose the correct matching:
1. A
2. B
3. C
4. D
Q36MCQ2 Marks·Choose one option
Asked 3 times Key ↓
1. A1→ iii, 2→ iv, 3→ i, 4 → ii
2. B1→ iii, 2→ i, 3 → ii, 4 → v
3. C1→ iii, 2→ iv, 3→ v, 4 → ii
4. D1→ iv, 2→ iii, 3→ i, 4 → ii
Q37MCQ2 Marks·Choose one option
Asked 2 times Key ↓
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. A6→ 1→ 4→ 5 → 2
2. B3→ 1→ 6→ 4 → 2
3. C3→ 5→ 1→ 4 → 2
4. D6→ 3→ 5→ 7 → 2
Q38MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. AOnly statement P is true
2. BOnly statement Q is true
3. CBoth statements are true
4. DNone of these statements is true
Q39MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A1→ vi, 2→ iii, 3 →v, 4 → i
2. B1→ i, 2→ iv, 3 → iii, 4→ ii
3. C1→ i, 2→ v, 3 → iii, 4→ iv
4. D1→ iv, 2→ iii, 3→ ii, 4 → i
Q40MCQ2 Marks·Choose one option
Asked 2 times Key ↓
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.
1. A5, 2, 1, 3, 4
2. B5, 1, 3, 2, 4
3. C1, 3, 2, 4, 5
4. D1, 5, 3, 2, 4
Q41MCQ2 Marks·Choose one option
Key ↓
1. ASigmoid of alignment scores
2. BReLU of alignment scores
3. CSoftmax of alignment scores
4. DLinear projections of alignment scores
Q42MCQ2 Marks·Choose one option
Asked 4 times Key ↓
Which one of the following statements is true?
1. AAttention mechanisms cannot be applied to the bidirectional RNN model
2. BAn image captioning network cannot be trained end-to-end even though we are using 2 different modalities to train the network
3. COne of the key components in the vanilla transformer are the recurrent connections that help them to deal with variable input length.
4. DNone of these
Q43MCQ2 Marks·Choose one option
Key ↓
1. A
2. B
3. C
4. D
Q44MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Which one of the following statements regarding hyperparameter tuning is false?
1. AValidation set is the best choice for tuning hyperparameters
2. BPicking hyperparameters that perform the best on the training set leads to overfitting
3. CTuning hyperparameters on the test set is an acceptable approach to find the most powerful model for generalization performance
4. DTest set performance, with hyperparameters tuned on the validation set, would be a realistic measure of how the model will generalize to new data.
Q45MCQ2 Marks·Choose one option
Asked 3 times Key ↓
1. A
2. B
3. C
4. D
Q46MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A1→ iii, 2→ iv, 3→ ii, 4 → i
2. B1→ i, 2→ iv, 3 → iii, 4→ ii
3. C1→ iv, 2→ iii, 3→ ii, 4 → i
4. D1→ iii, 2→ i, 3 → ii, 4 →v
Q47MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A1→ iv, 2→ v, 3→ i, 4 → ii
2. B1→ iii, 2→ i, 3 → v, 4→ ii
3. C1→ iv, 2→ iii, 3→ ii, 4 → i
4. D1→ iii, 2→ i, 3 → ii, 4→ v
Q48MCQ2 Marks·Choose one option
Asked 2 times Key ↓
What makes the Segment Anything Model (SAM) particularly advantageous for image annotation tasks compared to conventional segmentation approaches?
1. AIt has perfect segmentation accuracy
2. BIt has minimal hardware requirements for operation
3. CIt can generalize to segment objects it wasn’t specifically trained on
4. DIt automatically identifies and labels all objects in an image without user prompts
Q49MCQ2 Marks·Choose one option
Key ↓
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?
1. ABoth statements are correct, and Statement 2 correctly explains Statement 1.
2. BBoth statements are correct, but Statement 2 does not correctly explain Statement 1.
3. CStatement 1 is correct, but Statement 2 is incorrect.
4. DStatement 1 is incorrect, but Statement 2 is correct.
5. EBoth statements are incorrect.
Q50MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Why does DETR typically exhibit poor performance in detecting small objects compared to larger ones?
1. AThe CNN backbone used in DETR has a fixed receptive field that’s too large for small objects
2. BThe global self-attention mechanism in DETR tends to dilute the signal from small objects across all image locations
3. CDETR’s loss function actively discards all detections smaller than 50x50 pixels
4. DThe object queries in DETR are programmed to ignore any object smaller than 10
5. EDETR’s encoder-decoder architecture was specifically designed to only detect large objects
Q51MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following statements are false? (Select all that apply)
1. AGenerative adversarial networks (GANs) generate sharper images compared to Variational AutoEncoders (VAE)
2. BGAN is an example of an implicit density estimation model
3. CFully connected layers in mapping network of Style-GAN do not change the dimension of its input
4. DThe generator and discriminator are always trained together in a GAN
Q52MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following statements are true? (Select all possible correct options)
1. AAutoencoder are equivalent to Principal Component Analysis (PCA) provided we don’t use of non-linear activation functions
2. BWhen using global attention on temporal data, alignment weights are learnt for encoder hidden representations for all time steps
3. CPositional encoding is an important component of the transformer architecture as it conveys information about order in a given sequence
4. DIt is not possible to generate different captions for the same image that have similar meaning but different tone/style
5. EAutoencoders can not be used for data compression as its input and output dimensions are different
Q53MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following statements about CLIP are TRUE? (Select ALL that apply)
1. ACLIP uses a contrastive loss function to align text and image embeddings.
2. BCLIP requires image labels for training.
3. CCLIP is trained to maximize cosine similarity between embeddings of matching text-image pairs.
4. DCLIP model is trained with pretext tasks such as rotation prediction and inpainting to learn effective feature representations.
5. ETo use the CLIP model for image classification, it is imperative to finetune it on downstream datasets.
Q54MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which one of the following statements is true:
1. AWeight 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.
2. BL1 norm tends to create more sparse weights than L2 norm.
3. CDuring 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.
4. DA single McCulloch-Pitts neuron is capable of modeling AND, OR, XOR, NOR, and NAND functions.
Q55MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following are examples of a high-pass filter?
1. A
2. B
3. C
4. D
Q56MSQ2 Marks·Choose all correct options
Key ↓
Which of the following statements are true? (Select all that apply)
1. A
2. B
3. C
4. D
Q57MSQ2 Marks·Choose all correct options
Asked 3 times Key ↓
Which of the following techniques help control the exploding or vanishing gradient problem in recurrent neural networks?
1. AGradient clipping
2. BUse ReLU activation funtion to control exploding gradient and sigmoid activation to control vanishing gradients
3. CUse Long Short-Term Memory (LSTM)
4. DUse data augmentation methods (for example, reversing the input sequence)
Q58NAT2 Marks·Write the number
Asked 3 times Key ↓
Your answer:
Q59NAT2 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q60NAT2 Marks·Write the number
Asked 3 times Key ↓
Your answer:
Q61NAT2 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q62NAT2 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q63NAT2 Marks·Write the number
Asked 3 times Key ↓
Your answer:
Read the passage below, then answer Q64–Q65
Q64NAT1 Mark·Write the number
Asked 4 times Key ↓
Your answer:
Q65NAT1 Mark·Write the number
Asked 4 times Key ↓
Your answer:
Read the passage below, then answer Q66–Q69
Based on the above data, answer the given subquestions.
Q66NAT1 Mark·Write the number
Key ↓
Element 1: _______________
Your answer:
Q67NAT1 Mark·Write the number
Key ↓
Element 2: _______________
Your answer:
Q68NAT1 Mark·Write the number
Key ↓
Element 3: _______________
Your answer:
Q69NAT1 Mark·Write the number
Key ↓
Element 4: _______________
Your answer:
Read the passage below, then answer Q70–Q75
Based on the above data, answer the given subquestions.
Q70NAT1 Mark·Write the number
Asked 13 times Key ↓
Sigmoid _______________
Your answer:
Q71NAT1 Mark·Write the number
Asked 13 times Key ↓
Linear _______________
Your answer:
Q72NAT1 Mark·Write the number
Asked 13 times Key ↓
Indicator Function _______________
Your answer:
Q73NAT1 Mark·Write the number
Asked 13 times Key ↓
Softplus _______________
Your answer:
Q74NAT1 Mark·Write the number
Asked 13 times Key ↓
ReLu _______________
Your answer:
Q75NAT1 Mark·Write the number
Asked 13 times Key ↓
Leaky-ReLu _______________
Your answer:
Read the passage below, then answer Q76–Q86
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.
Q76NAT0.55 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q77NAT0.55 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q78NAT0.55 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q79NAT0.55 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q80NAT0.55 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q81NAT0.55 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q82NAT0.54 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q83NAT0.54 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q84NAT0.54 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q85NAT0.54 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q86NAT0.54 Marks·Write the number
Asked 22 times Key ↓
Your answer:

End Term · 31 Aug 2025 (AN)End Term

Q87MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Why does DETR typically exhibit poor performance in detecting small objects compared to larger ones? Choose the best answer in your opinion.
1. AThe CNN backbone used in DETR has a fixed receptive field that’s too large for small objects.
2. BThe global self-attention mechanism in DETR tends to dilute the signal from small objects across all image locations.
3. CDETR’s loss function actively discards all detections smaller than 50 × 50 pixels.
4. DThe object queries in DETR are programmed to ignore any object smaller than 10% of the image size.
5. EDETR’s encoder-decoder architecture was specifically designed to only detect objects larger than a cat.
Q88MCQ2 Marks·Choose one option
Key ↓
Consider the following Pytorch code: 1. x = torch.tensor([[-3.,2.,3.],[4.,-5.,7.],[20.,3.,2.]]) 2. print(torch.mean(x,dim=0,keepdim=True)) The output of the code is:
1. Atensor([ 7., 0., 4.])
2. Btensor([[ -7., 10., -4.]])
3. Ctensor([[ -7.],[7.],[4.]])
4. Dtensor([ -7., 7., 3.])
5. Etensor([[7., 0., 4.]])
6. Ftensor([[ 0.],[7.],[4.]])
Q89MCQ2 Marks·Choose one option
Key ↓
Which one of the following statements is false?
1. ALinear contrast stretching is a point operation.
2. BMoving average is an example of a local operation.
3. CConvolution in the spatial domain cannot be obtained through addition in the frequency domain.
4. DAll of these.
Q90MCQ2 Marks·Choose one option
Key ↓
What is the correct order of operations for processing an image through a Vision Transformer (ViT)?
1. AImage patching → Positional embedding → Linear projection of flattened patches → Transformer encoder→ Classification head
2. BImage patching → Linear projection of flattened patches → Positional embedding → Transformer encoder→ Classification head
3. CPositional embedding → Image patching → Linear projection of flattened patches → Transformer encoder→ Classification head
4. DLinear projection of Images → Image patching → Positional embedding → Transformer encoder → Classification head
5. ELinear projection of Images → Image patching → Positional embedding → Transformer encoder → Transformer decoder→ Classification head
Q91MCQ2 Marks·Choose one option
Key ↓
Vector Quantized Variational Autoencoder (VQ-VAE) utilizes a discrete latent representation as opposed to continuous latent spaces used in traditional VAEs. One of the key components of a VQ- VAE is the codebook, which consists of a set of learnable vectors. What is the primary role of the codebook in VQ-VAE?
1. AIt performs the non-linear transformation of the input data to a higher dimensional space.
2. BIt regularizes the encoder by penalizing complex encodings.
3. CIt provides a finite set of vectors which the encoder’s outputs are mapped to, effectively quantizing the latent space.
4. DIt decodes the quantized vectors back into the reconstructed input space.
Q92MCQ2 Marks·Choose one option
Key ↓
Given two normalized embeddings from CLIP, I = [0.2, 0.5, 0.3, 0.4] for an image and T = [0.1, 0.6, 0.3, 0.7] for a text description, compute the CLIPScore(I,T).
1. A− 69%
2. B0%
3. C69%
4. D50%
Q93MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Which one of the following statements regarding hyperparameter tuning is false?
1. AValidation set is the best choice for tuning hyperparameters
2. BPicking hyperparameters that perform the best on the training set leads to overfitting
3. CTuning hyperparameters on the test set is an acceptable approach to find the most powerful model for generalization performance
4. DTest set performance, with hyperparameters tuned on the validation set, would be a realistic measure of how the model will generalize to new data.
Q94MCQ2 Marks·Choose one option
Asked 4 times Key ↓
Which one of the following statements is false? (Pick the most appropriate one.)
1. AAttention mechanisms can be applied to the bidirectional RNN model
2. BAn image captioning network can be trained end-to-end even though we are using 2 different modalities to train the network
3. COne of the key components in the vanilla transformer is the recurrent connection that help them to deal with variable input length.
4. DAll of these.
Q95MCQ2 Marks·Choose one option
Asked 3 times Key ↓
You are designing an edge detection pipeline for noisy grayscale images. For each of the following goals, choose the most appropriate filter or method: 1. Smooth out noise before any edge detection step 2. Detect edges by finding zero-crossings in the second derivative 3. Highlight regions with a strong intensity gradient 4. Perform basic edge detection with a simple directional gradient filter
1. A1 → Gaussian filter, 2 → Laplacian of Gaussian, 3 → First derivative of Gaussian, 4 → Sobel operator
2. B1 → Sobel operator, 2 → Gaussian filter, 3 → Laplacian of Gaussian, 4 → First derivative of Gaussian
3. C1 → First derivative of Gaussian, 2 → Sobel operator, 3 → Gaussian filter, 4 → Laplacian of Gaussian
4. D1→Laplacian of Gaussian, 2→First derivative of Gaussian, 3→Sobel operator, 4→Gaussian filter
Q96MCQ2 Marks·Choose one option
Asked 3 times Key ↓
1. A
2. B
3. C
4. D
Q97MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A1→ iii, 2→ iv, 3→ ii, 4 → i
2. B1→ i, 2→ iv, 3 → iii, 4→ ii
3. C1→ iv, 2→ iii, 3→ ii, 4 → i
4. D1→ iii, 2→ i, 3 → ii, 4 →v
Q98MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A1→ iv, 2→ v, 3→ i, 4 → ii
2. B1→ iii, 2→ i, 3 → v, 4→ ii
3. C1→ iv, 2→ iii, 3→ ii, 4 → i
4. D1→ iii, 2→ i, 3 → ii, 4→ v
Q99MCQ2 Marks·Choose one option
Asked 2 times Key ↓
In which one of the following applications would you use a many- to-one RNN architecture? Choose the most appropriate answer.
1. AInput is a review of a movie and output is the sentiment of the given review
2. BInput is an image and output is the textual description of the given image
3. CInput is an image and output is the class to which the given image belongs
4. DInput is a sentence in English language and output is the same sentence translated to Hindi language
Q100MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Why might Segment Anything (SAM) be particularly useful in data annotation tasks compared to traditional segmentation models?
1. AIt produces perfect segmentation masks.
2. BIt can adapt to segment any object, even those not seen during training
3. CIt requires less computational resources
4. DIt automatically labels all objects in an image without user input
Q101MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following statements are true? (Select all possible correct options)
1. AVariational AutoEncoders (VAE) generate sharper images compared to Generative adversarial networks (GANs)
2. BGAN is an example of an implicit density estimation model
3. CFully connected layers in mapping network of Style-GAN change the dimension of its input
4. DIn a GAN, the generator and discriminator are trained in an alternating fashion, each using the other’s output to improve.
Q102MSQ2 Marks·Choose all correct options
Key ↓
Classifier-free guidance is a technique used in diffusion models to improve sample quality without the explicit use of a classifier. This technique involves modifying the sampling process based on a control parameter. If the control parameter, denoted as γ, is set to zero, what effect does this have on the generation process?
1. AIt fails to generate realistic samples
2. BIt removes all guidance, effectively making the process equivalent to the unconditional generation.
3. CIt maximizes the influence of the classifier, leading to highly detailed generations.
4. DIt can lead to more diverse samples compared to higher values of γ, as the generation process is less constrained by the conditional information.
Q103MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following statements are true? (Select all possible correct options)
1. AAutoencoder are equivalent to Principal Component Analysis (PCA) provided we don’t use of non-linear activation functions
2. BWhen using global attention on temporal data, alignment weights are learnt for encoder hidden representations for all time steps
3. CPositional encoding is an important component of the trans- former architecture as it conveys information about order in a given sequence
4. DIt is not possible to generate different captions for the same image that have similar meaning but different tone/style
5. EAutoencoders can not be used for data compression as its input and output dimensions are different
Q104MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following statements about CLIP are TRUE? (Select ALL that apply)
1. ACLIP uses a cross-entropy loss function to align text and image embeddings.
2. BCLIP requires image labels for training.
3. CCLIP is trained to maximize cosine similarity between embeddings of matching text-image pairs.
4. DCLIP model is trained with pretext tasks such as rotation prediction and inpainting to learn effective feature representations.
5. ETo use the CLIP model for image classification, it is imperative to finetune it on downstream datasets.
Q105MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which one of the following statements is true:
1. AWeight 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.
2. BL1 norm tends to create more sparse weights than L2 norm.
3. CDuring 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.
4. DA single McCulloch-Pitts neuron is capable of modeling AND, OR, XOR, NOR, and NAND functions.
Q106MSQ2 Marks·Choose all correct options
Asked 2 times Key ↓
Which of the following are examples of a high-pass filter?
1. A
2. B
3. C
4. D
Q107NAT2 Marks·Write the number
Asked 3 times Key ↓
Your answer:
Q108NAT2 Marks·Write the number
Asked 3 times Key ↓
Your answer:
Q109NAT2 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q110NAT2 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q111NAT2 Marks·Write the number
Asked 2 times Key ↓
Your answer:
Q112NAT2 Marks·Write the number
Asked 3 times Key ↓
What is the size of the feature map after applying two successive convolution operations with given parameters? Image size = 72×72, Kernel size = 5×5, Padding = 2 and Stride = 3. (In calculation, take floor(x) whenever x is non-integer.If the answer is N × N, write just one value N in the blank) __________________ .
Your answer:
Read the passage below, then answer Q113–Q114
Based on the above data, answer the given subquestions.
Q113NAT1 Mark·Write the number
Asked 4 times Key ↓
Your answer:
Q114NAT1 Mark·Write the number
Asked 4 times Key ↓
Your answer:
Read the passage below, then answer Q115–Q118
Based on the above data, answer the given subquestions.
Q115NAT2 Marks·Write the number
Key ↓
Element 1:___________
Your answer:
Q116NAT2 Marks·Write the number
Key ↓
Element 2:___________
Your answer:
Q117NAT1 Mark·Write the number
Key ↓
Element 3:___________
Your answer:
Q118NAT1 Mark·Write the number
Key ↓
Element 4:___________
Your answer:
Read the passage below, then answer Q119–Q129
Consider the BLIP model architecture with an input (image-text pair) as shown below:
For each of the entities given, enter the appropriate number from the figure that corresponds to it:
Based on the above data, answer the given subquestions.
Q119NAT1 Mark·Write the number
Asked 22 times Key ↓
Self-Attention _______________
Your answer:
Q120NAT0.5 Marks·Write the number
Asked 22 times Key ↓
BERT Text Encoder Bi Self Attention:__________________
Your answer:
Q121NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Image-Text Matching Loss (LITM ):__________________
Your answer:
Q122NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Image Grounded Text Encoder Bi Self Attention: ___________________
Your answer:
Q123NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q124NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Image Grounded Text Encoder Cross Attention: _________________
Your answer:
Q125NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Image-Text Contrastive Loss (LITC): ____________
Your answer:
Q126NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q127NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Causal Self Attention: _____________
Your answer:
Q128NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Your answer:
Q129NAT0.5 Marks·Write the number
Asked 22 times Key ↓
Image Grounded Text Decoder Cross Attention: _____________
Your answer:
Read the passage below, then answer Q130–Q135
Based on the above data, answer the given subquestions.
Q130NAT1 Mark·Write the number
Asked 13 times Key ↓
Sigmoid ______________
Your answer:
Q131NAT1 Mark·Write the number
Asked 13 times Key ↓
Linear ______________
Your answer:
Q132NAT1 Mark·Write the number
Asked 13 times Key ↓
Indicator Function ______________
Your answer:
Q133NAT1 Mark·Write the number
Asked 13 times Key ↓
Softplus _____________
Your answer:
Q134NAT1 Mark·Write the number
Asked 13 times Key ↓
ReLU _______________
Your answer:
Q135NAT1 Mark·Write the number
Asked 13 times Key ↓
Leaky-ReLU _____________
Your answer:

Quiz 2 · 3 Aug 2025 (AN)Quiz 2

Q136MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A
2. B
3. C
4. D
Q137MCQ2 Marks·Choose one option
Key ↓
Which of the following architectures introduced the ROI pooling layer for extracting fixed-size feature maps from variable-size region proposals?
1. AR-CNN
2. BFast R-CNN
3. CFaster R-CNN
4. DYOLO
Q138MCQ2 Marks·Choose one option
Key ↓
Which object detection method first integrated a Region Proposal Network (RPN) to generate region proposals within the detection model?
1. AR-CNN
2. BFast R-CNN
3. CFaster R-CNN
4. DYOLO
Q139MCQ2 Marks·Choose one option
Key ↓
Which architecture performs object detection as a single regression problem from image pixels to bounding boxes and class probabilities?
1. AR-CNN
2. BFast R-CNN
3. CFaster R-CNN
4. DYOLO
Q140MCQ2 Marks·Choose one option
Key ↓
1. A
2. B
3. C
4. D
Q141MCQ2 Marks·Choose one option
Asked 2 times Key ↓
Consider the following statements P and Q regarding AlexNet and choose the correct option: (P) In AlexNet, a trainable Local Response Normalization Layers were introduced to emulate the competitive nature of real neurons, where highly active neurons suppress the activity of neighboring neurons, creating competition among different kernel outputs. (Q) In AlexNet, a total of 8 Convolutional layers contain only about 6-7% of the total parameters hence account for the least computation.
1. AOnly statement P is true
2. BOnly statement Q is true
3. CBoth statements are true
4. DNone of these statements is true
Q142NAT2 Marks·Write the number
Key ↓
Consider two bounding boxes in an image: - Box A: top-left at (0, 0), bottom-right at (12, 12) - Box B: top-left at (4, 4), bottom-right at (16, 16) What is the Intersection over Union (IoU) between these two boxes ____________(in %)
Your answer:
Q143MCQ2 Marks·Choose one option
Asked 4 times Key ↓
Which one of the following statements is True?
1. AAttention mechanisms can be applied to the bidirectional RNN model
2. BAn image captioning network cannot be trained end-to-end even though we are using 2 different modalities to train the network
3. COne of the key components in the vanilla transformer are the recurrent connections that help them to deal with variable input length.
4. DAll of these.
Q144MCQ2 Marks·Choose one option
Key ↓
1. A
2. B
3. C
4. D
Q145MCQ2 Marks·Choose one option
Asked 2 times Key ↓
1. A
2. B
3. C
4. D
Q146MCQ2 Marks·Choose one option
Key ↓
1. A
2. B
3. C
4. D
Q147MCQ2 Marks·Choose one option
Asked 3 times Key ↓
Which of the following techniques help mitigate the vanishing gradient problem in recurrent neural networks? 1. Gradient clipping 2. Use of ReLU activation to preserve gradient flow 3. Use of LSTM units 4. Data augmentation (e.g., reversing input sequence)
1. A1 and 2 only
2. B2 and 3 only
3. C3 and 4 only
4. D1, 2, and 3 only
Q148MCQ2 Marks·Choose one option
Asked 3 times Key ↓
Which of the following techniques help mitigate the exploding gradient problem in recurrent neural networks? 1. Gradient clipping 2. Use of Sigmoid activation to prevent gradient growth 3. Use of LSTM units 4. Data augmentation (e.g., reversing input sequence)
1. A1 and 2 only
2. B1 and 3 only
3. C1, 2, and 3
4. DAll of these
Q149MCQ2 Marks·Choose one option
Asked 4 times Key ↓
Which one of the following statements is false? (Pick the most appropriate one.)
1. AAttention mechanisms cannot be applied to the bidirectional RNN model
2. BAn image captioning network cannot be trained end-to-end even though we are using 2 different modalities to train the network
3. COne of the key components in the vanilla transformer is the recurrent connection that help them to deal with variable input length.
4. DAll of these
Q150MSQ3 Marks·Choose all correct options
Key ↓
Which of the following is true regarding Hard Attention and Soft Attention?
1. ASoft Attention is smooth and differentiable
2. BVariance reduction techniques are used to train Soft Attention models
3. CSoft Attention is computationally cheaper than Hard Attention when the source input is large
4. DThe inference (test time) overhead is low in Hard Attention when compared to Soft Attention models
Q151MSQ3 Marks·Choose all correct options
Key ↓
Which of the following are true:
1. A
2. B
3. C
4. D
Q152NAT2 Marks·Write the number
Key ↓
Consider a Faster R-CNN object detection framework, which is being trained with a mini-batch size of 2. The RPN produces 512 proposals, out of which 256 and 128 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.
Your answer:
Read the passage below, then answer Q153–Q154
Suppose you want to build an encoder-decoder recurrent model with attention. The context vector at time step t is given by:
where α_(tj) refers to the alignment score between the decoder state at time t _____________and the encoder state at time _____________. Note: Please write your answer using variables
Based on the above data, answer the given subquestions.
Q153NAT1 Mark·Write the number
Key ↓
Decoder state at time t : _______________
Your answer:
Q154Short answer1 Mark·Write the answer
Key ↓
Encoder state at time: _______________
Your answer:
Read the passage below, then answer Q155–Q160
Consider a Gated Recurrent Unit (GRU) cell at time step ’t’.
For each of the entities in the given subquestions, enter the appropriate number from the figure that corresponds to it:
Based on the above data, answer the given subquestions.
Q155NAT0.66 Marks·Write the number
Key ↓
Hidden state h_(t−1)______________
Your answer:
Q156NAT0.66 Marks·Write the number
Key ↓
Hidden state h_t_____________
Your answer:
Q157NAT0.67 Marks·Write the number
Key ↓
Current Input x_t________________
Your answer:
Q158NAT0.67 Marks·Write the number
Key ↓
Update gate z_t_______________
Your answer:
Q159NAT0.67 Marks·Write the number
Key ↓
Reset gate r_t ______________
Your answer:
Q160NAT0.67 Marks·Write the number
Key ↓
New hidden state content e_h____________
Your answer:
Read the passage below, then answer Q161–Q162
Let input have size D_f × D_f × M where D_f = 10 and M = 256 and output feature map (after passing input through conv layer) has D_f × D_f × N size where N = 512. Assume padded convolution. Let width of the square kernel in conv layer be k where k = 7 (Ignore the bias term in the calculation). Calculate the number of parameters and computational cost for this convolution layer. Based on the above data, answer the given subquestions.
Q161NAT2 Marks·Write the number
Asked 2 times Key ↓
Number of Parameters:_____________
Your answer:
Q162NAT2 Marks·Write the number
Key ↓
Computational Cost:___________________
Your answer:
Read the passage below, then answer Q163–Q166
Let input have size D_f × D_f × M where D_f = 10 and M = 256 and output feature map (after passing input through conv layer) has D_f × D_f × N size where N = 512. Assume padded convolution. Let width of the square kernel in conv layer be k where k = 7 (Ignore the bias term in the calculation), calculate the number of parameters and computational cost, but make use of Depthwise Seperable convolution and pointwise convolution instead of standard convolution. Based on the above data, answer the given subquestions.
Q163NAT1 Mark·Write the number
Key ↓
Number of parameters for depthwise seperable convolution:_______________
Your answer:
Q164NAT1 Mark·Write the number
Key ↓
Computational Cost for depthwise seperable convolution:____________________
Your answer:
Q165NAT1 Mark·Write the number
Key ↓
Number of parameters for pointwise convolution:___________________
Your answer:
Q166NAT1 Mark·Write the number
Key ↓
Computational cost for pointwise convolution:__________________
Your answer:
Read the passage below, then answer Q167–Q171
Based on the above data, answer the given subquestions.
Q167NAT0.8 Marks·Write the number
Key ↓
Number of weights in Weight Matrix U1 is:______________
Your answer:
Q168NAT0.8 Marks·Write the number
Key ↓
Number of weights in Weight Matrix V1 is:______________
Your answer:
Q169NAT0.8 Marks·Write the number
Key ↓
Number of weights in Weight Matrix U2 is:__________________
Your answer:
Q170NAT0.8 Marks·Write the number
Key ↓
Number of weights in Weight Matrix V2 is:___________________
Your answer:
Q171NAT0.8 Marks·Write the number
Key ↓
Number of weights in Weight Matrix W is:_______________
Your answer:
Read the passage below, then answer Q172–Q179
Q172NAT0.5 Marks·Write the number
Key ↓
e_(21): _________________
Your answer:
Q173NAT0.5 Marks·Write the number
Key ↓
e_(22): _________________
Your answer:
Q174NAT0.5 Marks·Write the number
Key ↓
e_(23): _________________
Your answer:
Q175NAT0.5 Marks·Write the number
Key ↓
α_(21): _________________
Your answer:
Q176NAT0.5 Marks·Write the number
Key ↓
α_(22): _________________
Your answer:
Q177NAT0.5 Marks·Write the number
Key ↓
α_(23): _________________
Your answer:
Q178NAT0.5 Marks·Write the number
Key ↓
c_(21): _________________
Your answer:
Q179NAT0.5 Marks·Write the number
Key ↓
c_(22): _________________
Your answer:
Read the passage below, then answer Q180–Q187
Q180NAT0.5 Marks·Write the number
Key ↓
e_(21):_________________
Your answer:
Q181NAT0.5 Marks·Write the number
Key ↓
e_(22):_________________
Your answer:
Q182NAT0.5 Marks·Write the number
Key ↓
e_(23):_________________
Your answer:
Q183NAT0.5 Marks·Write the number
Key ↓
α_(21):_________________
Your answer:
Q184NAT0.5 Marks·Write the number
Key ↓
α_(22):_________________
Your answer:
Q185NAT0.5 Marks·Write the number
Key ↓
α_(23):_________________
Your answer:
Q186NAT0.5 Marks·Write the number
Key ↓
c_(21):_________________
Your answer:
Q187NAT0.5 Marks·Write the number
Key ↓
c_(22):_________________
Your answer: