Skip to content

Question 1 example 2

⬅️ [Question 2 example 1](<./Question 2 example 1.md>) | ⬆️ [Generated Test Questions](<./README.md>) | [Question 1 example 1](<./Question 1 example 1.md>) ➡️

Problem 1: Recall (25pts, 5pts each)
Directions: Use only the space provided. (Answer in max 1-2 sentences where applicable).

  1. How many degrees of freedom does a 2D Affine transformation have? [3, 4]

  2. True or False: In a perfectly parallel stereo camera setup, the disparity between corresponding points is directly proportional to the object's depth in the 3D scene. [5, 6]

  3. What specific hardware filter arrangement is used to capture RGB colors using a single camera sensor array, typically consisting of alternating filters with 2 Green, 1 Red, and 1 Blue per block? [7]

  4. When training a linear classifier or neural network, what is the primary purpose of applying L2 Regularization to the loss function? [8]

  5. In image segmentation using graphs, briefly explain why the Normalized Cut (Ncut) criterion divides the standard cut penalty by the volume (association) of the regions. [9, 10]


Solutions:

  1. 6 degrees of freedom. [3, 4]

  2. False. Disparity is inversely proportional to depth (objects closer to the camera move a greater distance across the visual field than objects far away) [5, 6].

  3. Bayer Filter (or Bayer Pattern). [7]

  4. It continuously pulls the weights down to penalize large weights, which encourages generalization and helps avoid overtraining. [8]

  5. It normalizes the inter-region cut to penalize degenerate, isolated single-node cuts, thereby encouraging more balanced segmentation sizes. [9-11]


⬅️ [Question 2 example 1](<./Question 2 example 1.md>) | ⬆️ [Generated Test Questions](<./README.md>) | [Question 1 example 1](<./Question 1 example 1.md>) ➡️