1

I am trying to implement "Real Time Head Pose Estimation fromConsumer Depth Cameras" by Fanelli et al. I need to train a random forest(regression) with the following criterion

  1. The predicted output is multivariate(yaw,pitch).
  2. The input features is four dimensional float type.
  3. The split function at each node is a custom function which needs to be computed at run-time using one of the features that has been randomly selected.
  4. The custom function needs to evaluated using a separate user created code snippet which might then be used by the random forest code.

My question is what random forest software frame work would support custom split functions and random feature selection for evaluating the custom function.

I have checked random forest implementations of scikit-learn, opencv, weka to begin with, but found that they don't satisfy my constraints

user27665
  • 121
  • 3
  • Feel free to implement it yourself. A random forest is just an ensembling of decision trees. you can check how to build a decision tree from scratch in https://www.youtube.com/watch?v=y6DmpG_PtN0&t=3s – Carlos Mougan Feb 18 '20 at 22:40

0 Answers0