Detecting Artificial Patterns
Hi!
I trained a classifier to detect a MacBeth colorchart which is an artificial pattern and should be easy to detect. Training was done detecting haarfeatures from 1500 negatives and 4000 positives using bg photos from google (http://tutorial-haartraining.googlecode.com/svn/trunk/data/negatives/) that were similar to the environment where the pattern is most probably to be found.
The statistics of the training seemed to be very reasonable:
It went through all 10-stages and produced a 33kb XML cascade description file.
However when I try to find the pattern, it detects all sorts of things:
Does anyone have an idea how to improve the settings or properly find that pattern?
hmm, cascade detection does not "see" colors at all.
if you want to do this for an AR application, maybe choose a pattern with more distinctive "shapes" ?
yea I'm just overlaying the results on top of the colored image. I see what you're saying about the shape, but it's not really for an AR application. It's more for autodetection of that colorchart to neutralize and calibrate the colors of the incoming image, so the pattern has to be this.
k,k. i see.
It's weird that the XML is only 33kb, cannot imagine 33kb are enough to properly describe the pattern.