Integrated Human Tracking Based on Video and Smartphone Signal Processing within the Arahub System

Abstract

Embedded platforms with GPU acceleration, designed for performing machine learning on the edge, enabled the creation of inexpensive and pervasive computer vision systems. Smartphones are nowadays widely used for profiling and tracking in marketing, based on WiFi data or beacon-based positioning systems. We present the Arahub system, which aims at integrating world of computer vision systems with smartphone tracking for delivering data useful in interactive applications, such as interactive advertisements. In this paper we present the architecture of the Arahub system and provide insight about its particular elements. Our preliminary results, obtained from reallife test environments and scenarios, show that the Arahub system is able to accurately assign smartphones to their bearers, based on visual and WiFi/Bluetooth positioning data. We show the commercial value of such system and its potential applications.

Introducing LogDL - Log Description Language for Insights from Complex Data

“…Methods that manipulate a LogDL description could take it from there and use those local features to induce/infer higher-level concepts. Works such as [38] demonstrate efficiency of such hybrid approach with respect to multimodal spatio-temporal data, whereby LogDL could be additionally used to express the domain knowledge of subject matter experts at that higher level of abstraction.…”