Abstract | ||
---|---|---|
A novel system for adapting generic trackers for long-term face tracking in unconstrained videos is proposed. The system treats the tracker as a black box. The only requirement is that the tracker can be reinitialized when needed. The system consists of a generic face detector trained offline and a validator trained online which helps to distinguish the target face from other people's faces and the background. We demonstrate this method on three state-of-the-art generic trackers: OpenTLD, Struck and MIL. For the experiments we use public face videos as well as our own dataset. In all our experiments our face tracking adaptation method shows superior results in comparison with the original trackers. |
Year | DOI | Venue |
---|---|---|
2015 | 10.1109/IVCNZ.2015.7761570 | 2015 International Conference on Image and Vision Computing New Zealand (IVCNZ) |
Keywords | Field | DocType |
generic trackers,face tracking,unconstrained videos,black box,face detection,OpenTLD,Struck,MIL | Black box (phreaking),BitTorrent tracker,Computer vision,Pattern recognition,Computer science,Artificial intelligence,Detector,Facial motion capture,Validator | Conference |
ISSN | ISBN | Citations |
2151-2191 | 978-1-5090-0358-7 | 0 |
PageRank | References | Authors |
0.34 | 10 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Maria Mikhisor | 1 | 1 | 1.02 |
Geoff Wyvill | 2 | 674 | 219.62 |
brendan mccane | 3 | 223 | 33.05 |
Steven Mills | 4 | 41 | 17.74 |