quote:Original post by Lithic
Since yann seems to not be interested in giving me an answer, do any of you others know anything about this?

It not a question about me not being interested, but more that I can''t monitor each thread 24/7, only waiting for a question directed at me.

To answer your question: yes, I use(d) neural nets to get the location. But I can''t really go into the details, because:

A) I didn''t write the NN, so I don''t know all the details.
B) Neural net theory is an extremely vast and complex topic, that fills entire libraries. You can''t simpy explain that in a forum post.

If you are interested in NN theory, and how to apply them to mathematical minimization and optimization problems, there is lots of good literature available. This subject undergoes very active research, and lots of online resources are also available. Ask in the AI forum about some good pointers.

Besides this, I would strongly recommend using a conventional approach to get the division plane instead. Writing a NN solely for that task is total overkill. We happened to have one lying around, that was written for a related problem a long time ago, and so we just plugged it in. But it has many drawbacks, especially performance requirements.