Object Recognition and Tracking for Smart Audio Guides

In this paper we address the problem of creating a smart audio guide that adapts to the actions and interests of tourists. Our guide performs automatic recognition of artworks and allows the users instant or deferred fruition of multimedia content. We use a compact CNN as computer vision system to back the whole application to performs object classification, localization and recognition. Tracking is used to improve the recognition accuracy over sequences of detections. We also provide an automatic pipeline for dataset creation based on the same tracking algorithm. The system, deployed on an NVIDIA Jetson TK1 and an NVIDIA Shield Tablet, has been tested in a real world environment.