A Graph Learning Based Framework for Billion-Scale Offline User Identification

Offline user identification is a scenario that users use their bio-information like faces as identification in offline venues, which has been applied in many offline scenarios such as verification in banks, check-in in hotels and making a purchase in offline merchants. In such a scenario, designing an identification approach to do extremely accurate offline user identification is critical. Most scenarios use faces to identify users and previous algorithms are mainly based on visual features and computer-vision models. However, due to the large variations such as pose, illumination and occlusions in offline scenarios, it remains a challenging problem for existing computer-vision algorithms to get a satisfying accuracy in real-world scenarios. Furthermore, billion-scale candidate users also require high efficiency and high accuracy for the approach.

In offline scenarios, users, venues and their different kinds of interactions can form a heterogeneous graph. Mining the graph can tell much information about users' offline habits and behaviors, which can be regarded as a great information supplement for user identification. In this paper, we elaborately design an offline identification framework considering two aspects. First, given a scanning face, we propose a 'local-global' retrieval mechanism to find one user from billion-scale candidate users. Second and most importantly, to make the verification between the scanning face, the retrieved user and the venue, we propose a novel Wide & Deep Based Graph Convolution Network to model both the visual information and the heterogeneous graph. Extensive offline and online A/B experimental results on a real-world industrial dataset demonstrate the effectiveness of our proposed approach. Nowadays, the whole approach has been deployed to serve billion-scale users to do offline identification in the industrial production environment.