CNN models working

its workinng but the approach is too blunt, gonna try another approach using landmarks to detect gestures that will make it more dynamic and robust.