1. The template-matching algorithm is not scale- and rotation-invariant by itself. How can we make it so for a) double the scale of the template image, and b) a 90-degrees-rotated version of the template image?
2. Use the GFTTDetector class to detect keypoints with the Harris corner-detection algorithm. You can set any values for the corner-detection algorithm.
3. The Hough transformation can also be used to detect circles in an image, using the HoughCircles function. Search for it in the OpenCV documentation and write a program to detect circles in an image.
4. Detect and draw the convex contours in an image.
5. Use the ORB class to detect keypoints in two images, extract their descriptors, and match them.
6. Which feature-descriptor-matching algorithm is incompatible with the ORB algorithm, and why?
7. You can use the following OpenCV functions and the given sample to calculate the time required to run any number of lines of code. Use it to calculate the time it takes for the matching algorithms on your computer:

double freq = getTickFrequency(); 
double countBefore = getTickCount(); 
 
// your code goes here .. 
 
double countAfter = getTickCount(); 
cout << "Duration: " << 
          (countAfter - countBefore) / freq << " seconds";