Click here for full text:
On the use of attention clues for an autonomous wearable camera
Pilu, Maurizio
HPL-2002-195R1
Keyword(s): saliency; wearable cameras; cameras; imaging; perception
Abstract: Autonomous wearable cameras that are able to capture moments by inferring situations of interest to their wearers might revolutionize the way people do and think of photography. The major technological challenge we face towards that goal is how we detect such situations. This report is my attempt to make a case for keeping the user in the loop in an attention detection framework against the conventional bottom-up computational approach of determining saliency from stimuli. A modification to the attention detection model of Barron-Cohen and an extension to the concept of deictic primitive put forward by Ballard et al. are proposed that better fit an autonomous wearable camera scenario.
19 Pages
Back to Index
| 
printable version
