Predicting Professions through Probabilistic Model under Social Context

In this paper, we investigate the problem of predicting people’s professions under social context. Previous work considering clothing information as well as fore/background context preliminarily proves the feasibility of predicting professions. In this paper, we discuss this problem in a more general case — multiple people in one photo with arbitrary poses, and argue that with appropriately built partial body features, spatial relations, and background context, more appealing results are achieved by a probabilistic model. We conduct experiments on 14 representative professions with over 7000 images, and demonstrate the model’s superiority with impressive results. Introduction In modern society, social status, connections, and people’s roles in a particular situation draw great attention since they are fundamental elements of daily life. To automatically recognize the social roles, researchers propose to determine single person’s demographical information by face at first, e.g., identity (Zhao et al. 2003), gender (Bourdev, Maji, and Malik 2011), age (Fu, Guo, and Huang 2010). Then more complex models considering pair-wise connections between people or context are introduced in (Naaman et al. 2005; Gallagher and Chen 2009; Wang et al. 2010; Berg et al. 2004; Xia et al. 2012). In this paper, we describe an application scenario that can potentially boost the performance of social characteristics analysis— parsing the professions of people in a photo. People tend to make friends with those of the same professions, and any social website could utilize this for friend recommendation or professional services. We argue that the professions in a photo can be more precisely parsed by social context in a probabilistic model. Our contributions First, we use poselet (Bourdev and Malik 2009; Bourdev et al. 2010), to capture the low-level feature, so we can deal with non-frontal upper body. Second, we use visual attributes (Farhadi et al. 2009; Kumar et al. ∗This research is supported in part by the NSF CNS award 1135660 and 1314484, Office of Naval Research award N0001412-1-0125 and N00014-12-1-1028, Air Force Office of Scientific Research award FA9550-12-1-0201, and IC Postdoctoral Research Fellowship award 2011-11071400006. Copyright c © 2013, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. (c) (a)