Someday, mobile robots will operate continually. Day  after day, they will be in receipt of a  never ending stream of images. In anticipation of this, this paper is about having a mobile robot generate apt and compact summaries of its life experience. We consider a robot moving around its environment both revisiting and exploring, accruing images as it goes. We describe how we can choose a subset of images which captures the essence of the robot’s visual experience – illustrating both what was ordinary and what was extraordinary – and which summarises the robot’s cumulative visual experience. Moreover we show how to do this such that the time cost of generating a summary is largely independent of the total number of images processed. No one day is harder to summarise than any other.

As the robot captures new images, they are encoded as a vector in topic space and organised in  a similarity topic graph using the star clustering algorithm. Images in a cluster possess similar visual topics, and cluster centres accurately capture the visual theme in the associated images(see figure opposite).

  • [PDF] R. Paul, D. Rus, and P. Newman, “How was your day? Online Visual Workspace Summaries using Incremental Clustering in Topic Space,” in Proc. IEEE International Conference on Robotics and Automation (ICRA), Minnesota, USA, 2012.
    [Bibtex]

    @inproceedings{PaulICRA2012,
    Address = {Minnesota, USA},
    Author = {Rohan Paul and Daniela Rus and Paul Newman},
    Booktitle = {Proc. IEEE International Conference on Robotics and Automation (ICRA)},
    Date-Modified = {2012-03-06 10:12:48 +0000},
    Keywords = {How Was Your Day?},
    Month = {May},
    Owner = {ashley},
    Pdf = {http://www.robots.ox.ac.uk/~mobile/Papers/2012ICRA_paul.pdf},
    Timestamp = {2012.02.02},
    Title = {How was your day? Online Visual Workspace Summaries using Incremental Clustering in Topic Space},
    Year = {2012}}

 

Screen Shot 2014-11-17 at 11.05.47

a) Buildings and trees

Screen Shot 2014-11-17 at 11.05.42

b) Parkland

 

Screen Shot 2014-11-17 at 11.03.17

c) Vehicles and the horizon

 

Figure: representative clusters from Oxford City Center and New College Data sets. Cluster centres represented in red