We seek to find all instances of a
specific person in a community contributed large photo collection. In
the above case, trained from a single image at the top-left, our system
correctly identifies 4 of the 5 correct matches shown above from a
collection of 282 images captures by 82 different people on a single
day at Trafalgar Square, London (Images downloaded from Flickr.com)
Abstract
Given a community-contributed set of photos of a crowded public event,
this paper addresses the problem of finding all images of each person
in the scene. This problem is very challenging due to large changes in
camera viewpoints, severe occlusions, low resolution and photos from
tens or hundreds of different photographers. Despite these challenges,
the problem is made tractable by exploiting a variety of visual and
contextual cues – appearance, timestamps, camera pose and
co-occurrence of people. This paper demonstrates an approach that
integrates these cues to enable high quality person matching in
community photo collections downloaded from Flickr.com
Paper
Supplementary Material
Poster [ppt | pdf]
Data: Coming Soon
Acknowledgements:
This work was supported in part
by National Science Foundation grants IIS-0811878, IIS-0963657 and IIS-0954083, the University of Washington
Animation Research Labs, Intel, Microsoft, and Google.
We are thankful to Flickr users whose photos we used.