This alternative visualisation is designed to be a rich and engaging means to browse a SenseCam photoset. The visualisation presents only a single representative image for each the most interesting events in a user’s photoset, allowing them to quickly review their activities.

The visualisation aggregates SenseCam images into discrete events each of which has been analysed to give a score of its interestingness. The most interesting events from throughout a user’s photoset are then displayed as an animated stream of events, each event represented by a keyframe image which on mouse-over plays back the full set of images as a video. Each event’s on screen movements and appearance are determined by its interestingness, duration, and time of the day at which it occurred. Temporal relationships between the events are maintained, so, events, which occur earlier in the day, will move onto the screen ahead of the later events. The interestingness of a particular event is visually indicated by the size and opacity of the keyframe image. Very interesting events are presented as large and solid items on screen, while the opposite applies to uninteresting events. The rate at which the event moves into view is also an indicator of its interestingness with more interesting events moving slower. After the event has moved into focus, it will briefly pause on screen. The length of which is related to its duration.

The visualization offers the ability to actively explore and perform directed searches or passively review the photo collection in a manner akin to a slideshow. By packaging the photos into events we have offered an effective way to reduce the amount of visual information that the user must navigate without impacting the richness of the dataset.

Future version of the visualization will seek to integrate additional sources of context information (e.g. GPS, Bluetooth information, computer activity) to provide automatic annotations of the events and enable search and relevance feedback within the interface. The visualisation seems well suited to the integration of SenseCam data from multiple users and this is something that we would also like to explore in future iterations. Slides