OpenCV_notes

Stuff that was mentioned in the meeting
	OpenCV uses BGR format. Basically just the opposite of *everything* else
	(Here's why: https://www.learnopencv.com/why-does-opencv-use-bgr-color-format/)
	
	OpenCV uses numpy arrays. If you're familiar with numpy, you'll use numpy matrices and can do whatever math you want with them
	When you're showing an image (to a person) the image has to be an array holding ints of type np.uint8, but you can use any array you want before that. So if I want to hold 16 bit values or average values of pixels, I can make an array of floats:
		myarray = np.zeros((Height, Width, Channels), dtype=np.float16)
	Then I can add something to make them outside of 0-255:
		myarray = myarray * 1.23456
	And that's fine as long as the shown image is the right type
		cv.imshow("image", myarray.astype(np.uint8))


Basic openCV functions:
	
	image = cv.imread("filename.png", mode)
		opens an image, probably won't be used
		modes are 0 (grayscale image), or 1 (color image)


	cv.imwrite("filename.png", array)
		saves a video frame as a png

	cap = cv.VideoCapture(source)
		Reads a video. The video can be a camera (which I think is usually 0) or a path to a file. cap is the "video object"
	
	ret, frame = cap.read()

		creates an array that is a frame of the video. every time you call cap.read() it will return the next frame
		ret is just if it failed, used for "if ret == badthings: break" stuff (end of the video)

	cv.imshow("name of window", array)
		shows an image in a window that is name "name of window"

		For videos, this will show a frame. To show video, you just use a while loop to alternate between:
			while(True):
				ret, frame = cap.read()
				cv.imshow("video",frame)

	cv.cvtColor(frame/array, cv.COLOR_THIS2THAT)
		converts one type of image to another. THIS2THAT can be a lot of things, but we will most likely use cv.BGR2GRAY most of the time, because certain functions require a grayscale image

	cv.threshold is important. I don't remember the exact syntax but it creates binary images. You can choose what value is used for the threshold (I can make an image so all pixels below a threshold are zero, and anything above the threshold is my choice of 1-255). You can also do it the opposite way (above the threshold is zero)

	We used cv.CascadeClassifier, which is what you'll likely be using. I don't know a ton about it though
		There's also a detection method called detectMultiscale, which might be relevant