Scalable Integration of Image and Face Based Augmented Reality
PubDate: August 2020
Teams: National University of La Matanza;National University of La Plata;CICPBA Scientific Research Commission of the Buenos Aires Province
Writers: Nahuel A. MangiaruaJorge S. IeracheMaría J. Abásolo
In this paper we present a scalable architecture that integrates image based augmented reality (AR) with face recognition and augmentation over a single camera video stream. To achieve the real time performance required and ensure a proper level of scalability, the proposed solution makes use of two different approaches. First, we identify that the main bottleneck of the integrated process is the feature descriptor matching step. Taking into account the particularities of this task in the context of AR, we perform a comparison of different well known Approximate Nearest Neighbour search algorithms. After the empirical evaluation of several performance metrics we conclude that HNSW is the best candidate. The second approach consists on delegating other demanding tasks such as face descriptor computation as asynchronous processes, taking advantage of multi-core processors.