Is import existing models for gesture recognition, motion image has been training well, who is a great god have action than in the core of the code?