A Review on Automatic Face-Name Association for Web Videos
Shweta Tadge, Dahake
Celebrity face labeling in web videos is huge and challenging task because of large deviation in the appearance of person or celebrity in the web videos. This work explores the problem of missing name and missing faces in unconstrained videos with user created metadata. Instead of depending upon supervised learning, a better relationship built from content of video, those relationship includes arrival of faces in different spatio-temporal contexts and visual similarities between faces. The knowledge base consist of weakly tagged images along with set of names and celebrity social networks. Merging of relationship along with knowledge base is carried out via conditional random field. Two types of face name association are investigated: within video face labeling and between video face labeling. The within video labeling takes care of noisy as well as incomplete labels in metadata, where null assignments for the names is permitted. Furthermore Between video face labeling addresses the flaws in metadata, particularly to correct incorrect names and label faces with missing names in the metadata of a video, by considering a gathering of socially associated videos for joint name inference.
Archived Files and Locations
|application/pdf 771.0 kB ||
access all versions, variants, and formats of this works (eg, pre-prints)