Implementation of two CNN models to classify the observations of a social greeting behavioral intervention. The first model is used to classify the video input captured by the camera of a NAO robot. The second classifies the audio coming from the microphones of the robot.