Whenever I talk about Total Recall, people ask: “won’t you get swamped by all that data?” It’s a great question. If it takes automatic capture to get close to logging everything, it will also take automatic organization, and, even more importantly, automatic summarization to deal with the scale of one’s lifelog. So, my Total Recall presentation isn’t complete without showing off the great work by Aiden Doherty, Alan Smeaton and others from Dublin City University (DCU).
The DCU SenseCam browser takes the several thousand pictures that a SenseCam can generate in a day, and produces a daily diary. The idea is to look for novelty as a proxy for interestingness. Faces and clothing patches are recognized, and they ask: are these the same old faces, or new ones? Looking accelerometer values, and taking location from a separate GPS into account, they can also distinguish movements that are out of the ordinary. The end result shows snapshots of your day, with the pictures of more interesting events taking up more space. When you put your mouse over an event’s picture, a time-lapse style video plays of all the pictures from that event.
Now the DCU SenseCam browser is being made public. Check out the SenseCam browser site for source code and links to more information about the project. This first public release doesn’t include the image comparison, but they have actually found they can get within a few percent of the same answers without it. But this release is just the start as they’ve got the capability to add in much more functionality, such as event importance or visual event search … stay tuned for more!