Question 22
Input feature map has M=64 channels. You apply depthwise 3×3 followed by pointwise 1×1 to produce N=128 channels. Ignore bias. What is the Total parameters?
Question 22: Input feature map has M=64 channels. You apply depthwise…
Show answer
Correct answer: 8768
Question 22 of 22 in the IIT Madras BS Deep Learning for Computer Vision (Deep Learning for Computer Vision) End Term paper sat on 10 May 2026, in the January 2026 term (Deep Learning For Computer Vision 06 May 26). It carries 4 marks.
More questions from this paper
- A standard RGB image of size 256×256 is stored as 8-bit unsigned integers per channel. What is the total memory (in byt…
- Which statement is correct?
- You apply a 5×5 filter with stride 1 and no padding on a 32×32 image. Output size is:
- The Sobel operator primarily estimates:
- SIFT is designed to be robust mainly to:
- In classical scale-space theory, increasing σ in Gaussian smoothing generally:
- Backpropagation is best described as:
- Which is a key effect of momentum?
- L2 regularization on weights most directly encourages:
- A key reason CNNs are parameter-efficient compared to fully connected nets for images is:
- Stacking multiple 3×3 convolutions (stride 1) increases receptive field because:
- In Faster R-CNN, the Region Proposal Network (RPN) outputs:
- Which statement is correct ?
- Vanishing gradients in vanilla RNNs occur mainly due to:
- Select all correct statements:
- Select all correct statements :
- Select all correct statements:
- Box A: top-left (0,0), bottom-right (8,8). Box B: top-left (4,4), bottom-right (12,12). Compute IoU as a decimal.
- For a single query, suppose similarity with the positive is s+=2 and with two negatives are s1=1, s2=0. InfoNCE loss: L…
- Input image: 64×64. Convolution: kernel 7×7, stride 2, padding 3. Compute output spatial size.
- A fully connected network maps input 100 → hidden 50 → output 10. Ignore bias. What is the total number of weights?