Hi list, i was wondering which is the best way of working with neural nets in openframeworks, i would like to train neural networks to recognize gestures that im getting using a kinect. Do anybody have tried something like this before?
In my opinion neural nets are best replaced with svm’s. There are a few things where they work well (like the endless character recognition examples), but they are non-deterministic (may or may not converge depending on the random initial thetas) and overall not as robust as svm in my experience.