Data robot

So far, we have watched over 400,000 hours of cricket videos in detail.  If all this were on DVDs, the shelf for the movie type boxes would have to be 2km long... (and this is just a subset of the 1.5 million hours of video we watched very quickly just to check if there were any crickets present)

Networks of video cameras will open new horizons in our understanding of ecosystems.  The main problem is not going mad watching all the footage. The best way to deal with this is to divide up the work, an alternative is to try to train a computer to do it.

Robot reading a cricket tag. Even with this poor resolution, the system identifies 1A

Robot tracking a cricket. Ellipses show 200 video frames as a cricket wanders around its burrow. The burrow entrance can be identified as central convergence point.

We are working to build software robots that can read the tag on a cricket's back and track its movement around a burrow.  Our aim is to build tools that other researchers can use on all sorts of animals. We're keen to collaborate to achieve this!