Project Proposal Machine learning with Weka for Kung fu- training purposes. Victoria Værnø
Feb 07, 2016
Project Proposal
Machine learning with Weka for Kung fu-training purposes.
Victoria Værnø
With what accuracy can the multilevel perceptron algorithm in Weka classify side-kicks?
Is it good enough for a beta training application for complete beginners?
DataCapture 500 side kicks of types:1. ”Bad Kick”2. ”Good Kick Guard Down”3. ”Good Kick Guard Up”
Attributes: Skeleton joint angles.
• Reduce movie sizes to 6 frames. • Make ARFF file with all the kicks. • Feed the ARFF file into Weka and experiment with the
algorithms.
Data Processing
Challenges • Skeleton recognition• Complex classes• Too much similarity which is
irrelevant to the classification
Questions?
Thoughts?
Tips?