Naaman, Mor and Garcia-Molina, Hector and Paepcke, Andreas and Yeh, Ron B. (2005) Leveraging Context to Resolve Identity in Photo Albums. Technical Report. Stanford.
Our system suggests likely identity labels for photographs in a personal photo collection. Instead of using face recognition techniques, the system leverages automatically-available context, like the time and location where the photos were taken. Based on time and location, the system automatically computes event and location groupings of photos. As the user annotates some of the identities of people in their collection, patterns of re-occurrence and co-occurrence of different people in different locations and events emerge. The system uses these patterns to generate label suggestions for identities that were not yet annotated. These suggestions can greatly accelerate the process of manual annotation. We obtained ground-truth identity annotation for four different photo albums, and used them to test our system. The system proved effective, making very accurate label suggestions, even when the number of suggestions for each photo was limited to five names, and even when only a small subset of the photos was annotated.
|Item Type:||Techreport (Technical Report)|
|Uncontrolled Keywords:||geo-referenced digital photos, face recognition, photo collections, context|
|Subjects:||Computer Science > Digital Libraries|
|Related URLs:||Project Homepage||http://www-diglib.stanford.edu/diglib/pub/|
|Deposited By:||Import Account|
|Deposited On:||30 Jan 2005 16:00|
|Last Modified:||22 Dec 2008 18:24|
Available Versions of this Item
- Leveraging Context to Resolve Identity in Photo Albums. (deposited 30 Jan 2005 16:00) [Currently Displayed]
Repository Staff Only: item control page