Annotation is important for personal photo collections because acquired metadata plays a crucial role in image management and retrieval. Bulk annotation, where multiple images are annotated at once, is a desired feature for image management tools because it reduces users' burden when making annotations.
The SAPHARI (Semi-Automatic PHoto Annotation and Recognition Interface) automatically creates meaningful photo clusters for efficient bulk annotation. It integrates automatically detected metadata into a bulk annotation interface where users can manually correct errors. The SAPHARI incorporates two automatic techniques; hierarchical event clustering and torso based human identification. Hierarchical event clustering provides multiple levels of "event" groups. For identifying people in photos, we introduce a new technique which uses clothing information rather than human facial features.
SAPHARI is built using HCIL's Piccolo.NET Toolkit for Zoomable User Interfaces.
Bongwon Suh, Graduate Research Assistant at HCIL, Ph.D. Candidate of Computer Science, University of Maryland
Ben Bederson, Director of HCIL, Assistant Professor of Computer Science, University of Maryland
Suh, B., and Bederson, B.B. (2007) Semi-Automatic Photo Annotation Strategies Using Event Based Clustering and Clothing Based Person Recognition, Interacting With Computers, Elsevier, 19 (4), 524-544.
Suh, B., and Bederson, B.B. (2004) Semi-Automatic Image Annotation Using Event and Torso Identification, Tech Report HCIL-2004-15, Computer Science Department, University of Maryland, College Park, MD
See the HCIL PhotoMesa project.
See the HCIL PhotoFinder project.