image processing - Working with decision trees -


i know tl;dr;

i'll try explain problem without bothering ton's of crappy code. i'm working on school assignment. have pictures of smurfs , have find them foreground background analysis. have decision tree in java has data (hsv histograms) 1 one single node. tries find best attribute (from histogram data) split tree on. executes split , creates left , right sub tree data split on both node-trees. data still kept in main tree able calculate gini index.

so after 26 minutes of analysing smurfs pc has giant tree splits , other data. question is, can give me global idea of how analyse new picture , determine pixels "smurf pixels". know have generate new array of data points hsv histograms of new smurf , need use generated tree determine pixels belong smurf.

can give me pointer on how this?

some additional information.
every decision tree object has split object has best attribute split on, value split on , gini index.

if need provide additional information i'd hear it.

ok. basically, in unoptimized pseudo-code: in order label pixels in new image:

for each pixel in new image:

  • calculate pixel's hsv features
  • recursively, starting tree's root :
  • is leaf? if is, give pixel dominant label of node.
  • otherwise, check splitting criterion against pixel's features, , go right or left child accordingly

i hope makes sense in context.


Comments

Popular posts from this blog

javascript - Enclosure Memory Copies -

php - Replacing tags in braces, even nested tags, with regex -