URGENT WORK Only OPENCV EXPERT Needed

Feature Maps

[Part A] Object Features- Using the Feature Type (3x2), show which of the following scaled views of the type are allowed using uniform scaling for a 6x6 window. Show your answer by coloring in the respective boxes of each candidate. Do not make any changes if the candidate map is not valid.

[Part B] For each of the feature maps that you selected in Part A, list how many checks of that map are searched in a 6x6 window. If the map is not valid simply mark an “X” in the count column.

In the 6x6 area (shown for reference only)

[Part C: Graduate Student] Assume the image is 640x480, how many 6x6 windows are there to check in the image? Show work by writing a Python script that takes into account any ImageWidth, ImageHeight, as well as any window size (WindowWidthxWindowHeight). The script function should return the number of evaluations and the locations using a list of coordinates of tuples (example: [(0,0), (0,1), (0,2)…]

[Part D: Graduate Student] Combine your answer in Part C (total of all checks) and Part B to calculate the total number of feature maps calculated.

AdaBoost

Use the following 2-d diagram to solve the AdaBoost problem.

9

8

7

6

5

4

3

2

FOR ANY TIES, choose the letter nearest the start of the alphabet, such as A before B or B before D.

[Part 1] Phase1: Assume that each point classified in error has an error score of 1.

[Part 2] Circle the best classifier (Phase1): ( A B C D )

[Part 3] Circle the points that the best classifier (Phase 1) had in error list: ( 1 2 3 4 5 6 7 8 9 )

[Part 4] Phase2: Assume that each point classified in error from the best classifier in Phase1 has an error score doubled for the next round. Re-calculate the total errors, but you do not need to re-score the best classifier line. Thus only 3 lines are given to you in the table:

[Part 5] Circle the best classifier (Phase 1): ( A B C D )

[Part 6] Circle the points that the best classifier (Phase 2) had in error list: ( 1 2 3 4 5 6 7 8 9 )

[Part 7] Phase3: Assume that each point classified in error from the best classifier in Phase3 has an error score double if the classifier incorrect labeled the point. Only errors are boosted. Re-calculate the total errors, but you do not need to re-score the best classifier line. Thus only 2 lines are given to you in the table: