Nov 15
aside
old business
ai-class.com bayes problem by hand ... several folks still missing this assignment.
So, let's look at it in class, eh?
current business
Today's homework : neural net demos to share with class?
new business
This week by request : computer vision (assignment coming)
resources
Please do look through these for Thursday's class.
- textbook chapter 24, Perception
- many wikipedia articles & the sources they reference, i.e.
discussion of basic ideas
computer/machine vision goals:
- feature detection (there's an edge there)
- object identification (that's a tree)
data types:
- black/white 2D pixel intensity
- stereoscopic
- video : feature "flows"
basic math concepts
- convolutions
- discrete convolution and "kernel"
- gaussian smoothing
- gradient
- higher order derivatives
typical steps (i.e. canny wikipedia article
- smooth image to remove noise; get to given scale
- compute gradient at each pixel; { |G| , theta (round to one of 4 angles, say)} from Gx, Gy
- search nearby for local max; get "thin edges"
- trace edges via high/low threshold; start high, trace along edge while above low
- output curves and/or corners ("points of interest")
OpenCV
Alex likes this with processing :
For next time:
- we will do the decision tree homework
- Alex demo of processing face recognition
- discuss computer vision readings
