Personal-location-based temporal segmentation of egocentric videos for lifelogging applications

Abstract Temporal video segmentation is useful to exploit and organize long egocentric videos. Previous work has focused on general purpose methods designed to deal with data acquired by different users. In contrast, egocentric video tends to be very personal and meaningful for the specific user who acquires it. We propose a method to segment egocentric video according to the personal locations visited by the user. The method aims at providing a personalized output and allows the user to specify which locations he wants to keep track of. To account for negative locations (i.e., locations not specified by the user), we propose a negative rejection method which does not require any negative sample at training time. For the experiments, we collected a dataset of egocentric videos in 10 different personal locations, plus various negative ones. Results show that the method is accurate and compares favorably with the state of the art.