Autotagging Facebook: Social network context improves photo annotation

Trevor Darrell,Todd Zickler,Zak Stone

doi:10.1109/cvprw.2008.4562956

Trevor Darrell, Todd Zickler + Show 1 more

Open Access

https://doi.org/10.1109/cvprw.2008.4562956

Copy DOI

Abstract

Most personal photos that are shared online are embedded in some form of social network, and these social networks are a potent source of contextual information that can be leveraged for automatic image understanding. In this paper, we investigate the utility of social network context for the task of automatic face recognition in personal photographs. We combine face recognition scores with social context in a conditional random field (CRF) model and apply this model to label faces in photos from the popular online social network Facebook, which is now the top photo-sharing site on the Web with billions of photos in total. We demonstrate that our simple method of enhancing face recognition with social network context substantially increases recognition performance beyond that of a baseline face recognition system.

Highlights

An increasing number of personal photographs are uploaded to online social networks, and these photos do not exist in isolation
We take a first step in this direction; we focus on the specific problem of automatic face recognition in personal photographs
Personal photos are highly variable in apperance but are increasingly shared online in social networks that contain a great deal of information about the photographers, the people who are photographed, and their various relationships

Summary

Introduction

An increasing number of personal photographs are uploaded to online social networks, and these photos do not exist in isolation. Each shared image likely arrives in a batch of related photos from a trip or event; these are associated with their photographer and broadcast out to that photographer’s hundreds of online friends, and they join a collection of billions of other photographs, some of which have been manually labeled with the people they contain and other information. Our investigation uses photos and context drawn from the online social network Facebook, which is currently the most popular photo-sharing site on the Web. With over 70 million active users, Facebook already hosts a collection of billions of personal photographs, and more than 14 million new photos are posted every day [5, 6]. Individuals are observed with a wide variety of expressions and poses, lighting varies tremendously, and other complications such as sunglasses and heavy makeup abound

Methods

Findings

Discussion

Conclusion