Quiz Space

Deep Learning Practice · End Term · 13 Sept 2026 · May 2026 term · Set 1

Question 28: Consider the multi-task loss used to train the original …

Question 28

+2 marksOne or more correct options

Consider the multi-task loss used to train the original YOLOv1 model. Which of the following statements about it are correct?

Select all that apply.

  1. A

    The loss predicts the square roots of width and height so that a fixed absolute error is penalized more for small boxes than for large boxes.

  2. B

    The hyperparameters are set to λ_coord = 5 (upweighting localization error) and λ_noobj = 0.5 (downweighting confidence error for cells with no object).

  3. C

    The classification term for a cell is only penalized when an object is actually present in that grid cell.

  4. D

    For the predictor "responsible" for an object, the target confidence value is fixed at 1, independent of the box's IoU with the ground truth.

Show answer

Correct answers

  • A

    The loss predicts the square roots of width and height so that a fixed absolute error is penalized more for small boxes than for large boxes.

  • B

    The hyperparameters are set to λ_coord = 5 (upweighting localization error) and λ_noobj = 0.5 (downweighting confidence error for cells with no object).

  • C

    The classification term for a cell is only penalized when an object is actually present in that grid cell.

See the whole paperTake it as a mock test

Question 28 of 31 in the IIT Madras BS Deep Learning Practice (Deep Learning Practice) End Term paper sat on 13 Sept 2026, in the May 2026 term (Deep Learning Practice 13 Sep 26 (Session 2)). It carries 2 marks.

More questions from this paper

  1. Q1Based on the above data, answer the given subquestions.
  2. Q2Based on the above data, answer the given subquestions.
  3. Q3Using the actual flattened feature size from previous question number 2, how many learnable parameters (weights + bias)…
  4. Q4Which of the following statements about the (corrected) model are true?
  5. Q5Figure question
  6. Q6Figure question
  7. Q7Which layer is responsible for the decoder failing to reach the target 120×120 output, and why?
  8. Q8Figure question
  9. Q9Figure question
  10. Q10Figure question
  11. Q11What is the parameter compression ratio (standard ÷ depthwise-separable), rounded to 2 decimal places?
  12. Q12Figure question
  13. Q13Which of the following statements correctly identify a real bug in the given code ?
  14. Q14Based on the above data, answer the given subquestions.
  15. Q15Based on the above data, answer the given subquestions.
  16. Q16A convolutional layer receives a feature map with 16 input channels and applies 32 filters, each of spatial size 5 × 5.…
  17. Q17A convolutional layer is applied to a 32 × 32 input using a 5 × 5 filter with zero-padding of 2 and stride 2. Using the…
  18. Q18The final classification head applies a softmax over the pre-activation scores (logits). For a 3-class problem, the log…
  19. Q19Following the VGG design philosophy of replacing a single large-kernel convolution with a cascade of small 3 × 3, strid…
  20. Q20Figure question
  21. Q21The decoder of a depth-estimation network upsamples feature maps using a transposed convolution ("up-convolution") with…
  22. Q22In the original U-Net used for dense prediction (and adaptable to depth estimation), each encoder stage applies two suc…
  23. Q23The SRGAN discriminator is a stack of 8 convolutional layers, all using 3 × 3 kernels with padding 1, and strides alter…
  24. Q24Efficient restoration networks (e.g., LaKDNet) replace standard convolutions with depth-wise separable convolutions (a …
  25. Q25Compared with the sigmoid activation, which of the following are correct reasons for preferring the ReLU activation in …
  26. Q26VGGNet replaces large convolution kernels with stacks of small 3 × 3 convolutions. Which of the following statements co…
  27. Q27Which of the following statements correctly describe the Region Proposal Network (RPN) in Faster R-CNN?
  28. Q29Regarding the channel-attention module in CBAM, which of the following are correct?
  29. Q30SRGAN performs photo-realistic super-resolution using a GAN. Which of the following statements are correct?
  30. Q31In the multi-scale deep network for single-image depth estimation (Eigen et al.), the architecture is split into a coar…