Presentation given at Personal Digital Archiving on promising technologies demonstrated in research literature that may ultimately improve annotation and management of personal digital photo collections.
Photographic prints from different time periods have different, identifiable color signatures based on the distinct succession of chemical processing technologies, facilitating EXIF timestamp generation.
The orientation of digitized photos can be detected and automatically corrected.
Cities can be identified by far less than their most visually iconic landmarks. Photos from different European cities were successfully localized relying on seemingly trivial elements in the visual environment: lamp-posts, grating iron-work, door frames, etc.
EXIF timestamps and the visual content of photos can be used to organize them into probable “events.”
Extending recognition algorithms to examine other parts of the body aside from the face increases accuracy.
Knowing which individuals are commonly in photos with which other individuals informs person recognition.
Who an individual is connected to and interacts with on social networks can help narrow likely candidates for person annotation in photos.
The angel from which the photo was shot relative to a vanishing point and the ground plane can be used to infer the height of the camera, which may provide insight into who the photographer was.
Muse is a personal data mining application for e-mail that shows, among other things, the frequency of communications with different individuals and groups of people over time; perhaps something like it could be adapted for use with photos?
A Survey of Research Prospects for more Manageable Personal Digital Photo Collections
A Survey of Research Prospects
for more Manageable Personal Digital Photo Collections Nicholas Taylor @nullhandlePersonal Digital ArchivingFebruary 21, 2013 “Photo Mosaic-Orange Daisy” by Flickr user BerniMartin under CC BY-ND 2.0
recognizing persons using body patch
matching Suh and Bederson: “Semi-Automatic Photo Annotation Strategies Using Event Based Clustering and Clothing Based Person Recognition”Cooray et al.: “Identifying Person Re-Occurrences for Personal Photo Management Applications”