Homework 2 Object Finder
Andrew Dunn ad2718
Prof Shree Nayar Computer Vision CS 4371 Fall 2011
resubmission 10-19-2011
p1 threshold 142 p3 no additional properties added to database p4 roundness and eMin used as criteria for recognition. roundness threshold - a difference of 0.02 eMin threshold - difference of 50000000 (this could be helpful for scaling, as images could have similar roundness, but scaled differently, therefore representing different objects).
note- I found the source of my error with the sequential labelling algorithm and equivalence set. I was processing labels directly on the image struct through setPixel() rather than in an external data vector. setPixel() only takes ints up to 255, and then it starts again 1. So, equivalences happened between the wrong numbers (such as 255 and 1 rather than 255 and 256), leading to one large equivalent set.
When I switched to an external storage grid for processing it worked fine.