Summer School 2013
![](../../2012/summerschool/images/1800px-Montafon_Panorama.jpg)
Photo by Böhringer
Classification and Regression Forests
by Prof. Dr. Jürgen Gall
A family of methods that can handle large amount of training data efficiently and that are inherently suited for multi-class problems are based on random forests.
Random forests are ensembles of randomized decision trees that can be optimized for classification, regression, or even combined classification and regression tasks.
The most prominent application of random forest is the detection of human body parts from depth data. The method was trained on 900k training examples to detect 31 body parts
(classes) and runs at around 200 frames per second on the Xbox GPU. This commercial application demonstrates the practicability of random forests for large-scale real-world computer vision problems.
References
Theory:Applications:
- http://research.microsoft.com/en-us/projects/vrkinect/
- http://www.iai.uni-bonn.de/~gall/projects/houghforest/houghforest.html
- http://www.iai.uni-bonn.de/~gall/projects/houghMC/houghMC.html
- http://www.iai.uni-bonn.de/~gall/projects/houghAR/houghAR.html/
- http://www.iai.uni-bonn.de/~gall/projects/RFhead/RFhead.html
- http://www.iai.uni-bonn.de/~gall/projects/facialfeatures/facialfeatures.html
- http://www.iai.uni-bonn.de/~gall/projects/humanpose2d/humanpose2d.html