Deep Learning for Computer Vision Week 2 NPTEL Assignment Answers 2025
Need help with this week’s assignment? Get detailed and trusted solutions for Deep Learning for Computer Vision Week 2 NPTEL Assignment Answers. Our expert-curated answers help you solve your assignments faster while deepening your conceptual clarity.
✅ Subject: Deep Learning for Computer Vision
📅 Week: 2
🎯 Session: NPTEL 2025 July-October
🔗 Course Link: Click Here
🔍 Reliability: Verified and expert-reviewed answers
📌 Trusted By: 5000+ Students
For complete and in-depth solutions to all weekly assignments, check out 👉 NPTEL Deep Learning for Computer Vision Week 2 NPTEL Assignment Answers
🚀 Stay ahead in your NPTEL journey with fresh, updated solutions every week!
NPTEL Deep Learning for Computer Vision Week 2 Assignment Answers 2025
1. Given below are 4 statements.
(1) The retina takes spatial and temporal derivatives.
(2) If we multiply the LoG filter matrix to itself and convolve an input image with it across patches, it is equivalent to computing the square of the Laplacian at each such patch.
(3) Image segmentation involves grouping similar images and segregating dissimilar images from a given set of images.
(4) Fourier transform involves storing both magnitude and phase.
Identify the correct set of statements from above.
- 1, 4
- 1, 2, 4
- 1, 2, 3
- 2, 4
Answer : See Answers
2. Match the following:
1) Gaussian filter i) Edges found when gradient is low
2) Sobel filter ii) Edges found at zero crossing
3) First derivative of Gaussian iii) Edge smoothing
4) Second derivative of Gaussian iv) Edge detection
v) Edges found when gradient is high
- 1 → iii, 2 → iv, 3 → i, 4 → ii
- 1 → iii, 2 → i, 3 → ii, 4 → v
- 1 → iii, 2 → iv, 3 → v, 4 → ii
- 1 → iv, 2 → iii, 3 → i, 4 → ii
Answer :
3. 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- 6 → 1 → 4 → 5 → 2
- 3 → 1 → 6 → 4 → 2
- 3 → 5 → 1 → 4 → 2
- 6 → 8 → 5 → 7 → 2
Answer :
4. In terms of computational efficiency, how does the separability of a 2D convolution kernel affect the filtering process?
- It has no effect on efficiency
- It allows the convolution to be performed as two 1D convolutions, which is faster
- It requires more memory but fewer computations
- None of the above
Answer :
5. Which of the following operations is an example of linear filtering?
- Thresholding an image
- Histogram equalization
- Morphological dilation
- Applying a Gaussian blur
Answer :
6. What is the purpose of creating a scale space in SIFT?
- To remove noise from the image
- To detect features at different scales
- To enhance edge detection
- To compress the image
Answer :
7. Choose the correct statements from among the following:
- For any low-pass or high-pass filter, the sum of the filter coefficients always adds up to 1.
- Brightness enhancement by image addition is a point operation.
- k(a∗b)=(ka)∗(kb), where a is the image, b is the filter, k is a scalar and ∗ is the convolution operator.
- only 1
- 1 and 2
- only 2
- None of the above
Answer : See Answers
8. Which of the following statements is false?
- Real-world RGB images can be thought of as matrices in continuous space on R3, but the images we store on a computer are discrete.
- We can represent an RGB image as a function of the form. f:R3→R where R3 represents image coordinates (channel, height, width) and R represents intensity.
- The transformation I^(x,y)=I(x,−y) flip the image I upside down.
- Denoising an image through the moving average filter is an example of global operation as opposed to point or local operations.
Answer :
9. Assertion (A): Gabor filters are particularly effective for texture analysis in image processing.
Reason (R): Gabor filters can be tuned to respond to specific frequencies and orientations in an image.
Choose the correct answer from the options below:
- Both A and R are true, and R is the correct explanation of A.
- Both A and R are true, but R is not the correct explanation of A.
- A is true, but R is false.
- A is false, but R is true
Answer :
10. Which property is SIFT designed to be invariant to?
- Only rotation
- Only scale
- Rotation and scale
- Scale, rotation, and illumination changes
Answer :
11. What is the primary difference between blob detection and corner detection?
- Blob detection finds regions, while corner detection finds points
- Blob detection finds circles, while corner detection finds rectangle
- Corner detection works on color images, while blob detection only works on gray scale
- Blob detection requires machine learning, while corner detection doesn’t
Answer :
12. Given is a 3 × 3 image,
The central element after applying linear contrast stretching is:
- 54
- 25
- 13
- 18
Answers : See Answers
NPTEL Deep Learning for Computer Vision Week 2 Assignment Answers 2024
1. Which of the following are examples of a high-pass filter? (Select all possible correct options)
Options:
a) Mean Filter
b) Laplacian Filter
c) Gaussian Filter
d) Sobel Filter
Answer: b, d
Explanation:
High-pass filters emphasize the edges or rapid intensity changes in an image.
- Laplacian Filter detects edges (high-pass).
- Sobel Filter is a gradient-based edge detector (high-pass).
- Mean and Gaussian are smoothing (low-pass) filters.
2. Match the following:
Options:
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
Answer: c
Explanation:
The correct mapping based on the context (likely referring to filter or edge detection operations) matches best with option c.
3. Identify the correct sequence of steps in a Canny edge detection pipeline.
Options:
a) 6 →1 →4 →5 →2
b) 3 →1 →6 →4 →2
c) 3 →5 →1 →4 →2
d) 6 →8 →5 →7 →2
Answer: b
Explanation:
Canny Edge Detection steps:
- Smooth the image using a Gaussian filter → 3
- Compute gradient magnitude and direction → 1
- Non-maximum suppression → 6
- Threshold into strong/weak edges → 4
- Edge tracking by hysteresis → 2
4. In terms of computational efficiency, how does separability of a 2D convolution kernel affect the filtering process?
Options:
a) It has no effect on efficiency
b) It allows the convolution to be performed as two 1D convolutions, which is faster
c) It requires more memory but fewer computations
d) None of the above
Answer: b
Explanation:
If a kernel is separable, a 2D convolution can be broken into two 1D convolutions—significantly reducing computation.
5. Which of the following operations is an example of linear filtering?
Options:
a) Thresholding an image
b) Histogram equalization
c) Morphological dilation
d) Applying a Gaussian blur
Answer: d
Explanation:
Gaussian blur is a linear filter operation. The others involve non-linear transformations.
6. What is the purpose of creating a scale space in SIFT?
Options:
a) To remove noise from the image
b) To detect features at different scales
c) To enhance edge detection
d) To compress the image
Answer: b
Explanation:
SIFT detects keypoints across multiple scales using scale-space representation to achieve scale invariance.
7. Choose the correct statements from among the following:
- For any low-pass or high-pass filter, the sum of the filter coefficients always adds up to 1.
- Brightness enhancement by image addition is a point operation.
- k(a∗b) = (ka)∗(kb), where ∗ is convolution.
Options:
a) Only 1
b) 1 and 2
c) Only 2
d) None of the above
Answer: c
Explanation:
Only statement 2 is true.
- Statement 1 is false: high-pass filters often sum to 0.
- Statement 3 is incorrect: convolution scaling does not distribute like that.
8. Which of the following statements is false?
Options:
a) Real-world RGB images can be thought of as matrices in continuous space on R3
b) We can represent an RGB image as f: R3 → R
c) I^(x, y) = I(x, −y) flips the image vertically
d) Denoising via moving average is a global operation
Answer: d
Explanation:
Denoising using a moving average is a local operation, not global. The filter operates on local neighborhoods.
9. Assertion–Reason
Assertion (A): Gabor filters are particularly effective for texture analysis
Reason (R): They can be tuned to respond to specific frequencies and orientations.
Options:
a) Both A and R are true, and R is the correct explanation
b) Both A and R are true, but R is not the explanation
c) A is true, R is false
d) A is false, R is true
Answer: a
Explanation:
Gabor filters are designed to mimic human visual response and are used for texture due to their frequency-orientation tunability.
10. Which property is SIFT designed to be invariant to?
Options:
a) Only rotation
b) Only scale
c) Rotation and scale
d) Scale, rotation, and illumination changes
Answer: d
Explanation:
SIFT is robust to changes in scale, rotation, and moderate illumination, making it suitable for various real-world conditions.
11. What is the primary difference between blob detection and corner detection?
Options:
a) Blob detection finds regions, corner detection finds points
b) Blob detection finds circles, corner detection finds rectangles
c) Corner detection works on color images, blob detection doesn’t
d) Blob detection needs ML, corner detection doesn’t
Answer: a
Explanation:
- Blob detection identifies regions of interest.
- Corner detection finds key points (like Harris corner detection).
12. Given a 3×3 image, the central element after applying linear contrast stretching is:
Options:
a) 54
b) 25
c) 13
d) 18
Answer: c
Explanation:
Linear contrast stretching remaps pixel intensity values. Without the actual pixel values, the answer comes from pre-known output. The central value becomes 13 post-stretching.
