EE5731: Q&A for CA1
Before asking questions to the GAs, please check the Q&As below. Basically you are not
allowed to use any existing high level functions (i.e., functions that call many other functions)
to complete the assignment. More detail is available below. If you have questions that are
not here, please contact the GAs.
1. Continuing on Q2 below, my implementation on cv.warpPerspectiv is very slow. It
works fine for Part 3, but for the rest of the parts in CA1, it will be too slow to
progress. Can I use my own implementation of image transformation only for Part 3,
and use cv.warpPerspective for the rest of CA1?
○ Yes, you are allowed. (Related to Q2 and Q7 below)
2. The Q&A below states that “Am I allowed to use an available function for image
transformation?” and the answer is “No, you are not allowed”. Does this mean that
we are not allowed to use cv.warpPerspectiv to transform image 1 to image 2
perspectives with our calculated homography? But I think cv.warpPerspectiv is the
warp function we need to implement in part 7, right? So you mean we need to
implement the warp function before part 7 since we need this function for part 1 to
part 6?
○ Yes, you’re right. (Related to Q1 and Q7)
3. Can I use numpy’s functions?
○ If you use numpy for basic matrix operations (like multiplication, subtraction,
etc) , then the answer is yes. You are also allowed to use the SVD available
function.
4. Can I use any available package for extracting SIFT?
○ Yes, the recommended package is from OpenCV
5. For part 1, what are the sizes of the Gaussian and Sobel kernels?
○ It's up to you to define the kernels including their sizes.
6. Can I use matlab's 2D convolution function?
○ No. Since, one of the purposes of this assignment is to know how to
implement the 2D convolution.
7. Am I allowed to use an available function for image transformation?
○ No, you are not allowed. Please read Q2 and Q1 above for further
information.
8. Do we need to manually write code for matching the descriptors or can we use
available functions?
○ You must write your own code.
9. Am I allowed to use OpenCV's basic functions for reading, writing, and displaying
images?
○ Yes
10. CA1 states 'You must write a report in a PDF format as instructed below (and indicate
the parts and Numbers clearly)'. So do we need to submit not only the Jupyter
notebook file, but also a report?
○ No separate report is required. The expectation is that all answers are shown
in the Jupyter notebook submission. I have revised the CA1 instructions.
11. About the requirements of code: Do we need to declare the version of the library we
used? E.g., OpenCV has different versions.
○ Yes, please declare the versions of any libraries you use.