The SenseCam is a passively capturing wearable camera that takes approximately 3,000 images on average per day. This provides a user with an extensive visual diary. Possible applications making use of this device include helping dementia sufferers recall events from short-term memory, and also this device can be used by tourists to maintain an extensive image collection of their trip. However a large image collection will quickly
build up, with an average of 1 million images captured each year. This presents a considerable challenge in terms of managing such a large collection and to make it accessible for users. Our Microsoft funded project investigates novel approaches in addressing this challenge.
We carry out the following processing:
Segment images into distinct events, e.g. having breakfast, talking to friend, being on bus, etc.
- Allow users the ability the search for similar events … e.g. when was the last time I was talking to my friend John?
- Given a number of events in a day we try to determine the most important event from that day, i.e. the most novel/unique event. Going to a football match will be a bigger highlight than have breakfast, as it occurs quite infrequently.
- Given an event, which (keyframe) image can we select as the image that best represents that event?. E.g. I can just look at 1 picture from the event, which will help me to best remember this event?
- We investigate numerous visualization techniques to best display this information back to the user, to give them a summary of their day/week/life.