notes.txt

2/10
-figure out a way to manually measure circles in graph
    -compare to output of sandbox.py to make sure that DoH is predicting properly
    -probably need 50-100 manually classified pictures from flickr


-classifier will predict the blob that represents a golf ball given a set of n blobs
****NOTE: COLOR MAY REFER TO A RANGE OF RGB VALUES********
    -n blobs will include attributes such as:
        -(x,y) center
        -radius
        -color that appears most frequently in blob
        -color that appears most frequently surrounding bottom 50% of blob
        -color that appears most frequently in surrounding top 50% of blob
        -percentage of blob covered by one contiguous color
        -more?

-save output from pipeline to ARFF file
    -ARFF FORMAT: http://www.cs.waikato.ac.nz/ml/weka/arff.html


2/14
-find more features to extract
-still need to write pipeline
-read PIL documentation to get various features already needed
    -can be done with ImageJ if not possible with pillow