This paper presents an approach to Soundpainting-based gesture recognition, using a machine learning framework developed in Python. This project builds upon our previous work, which focused on incorporating computational musical agents into an electroacoustic performance ensemble. In so doing, the system enables direct control of machine agents via gesture recognition. While specifically designed for thee Doug Van Nort Electro-Acoustic Orchestra (EAO), our approach offers flexible training and model generation for other composers and Soundpainting environments. At the time of writing the system is capable of detecting 41 Soundpainting gestures, and is modular and extensible to allow for expansion of the gesture dictionary. Gesture recognition is completed using four separate models, each targeting a separate vertical slice of the Soundpainter. Additionally, each gesture is broken down into “gesture components” (one or two combined individual hand gestures), which taken together make up a full combination gesture. Technical details and system design justifications are presented for model training, gesture recognition, and accompanying Max-based control patches. Initial testing results are presented.