The LVis or Leabra Vision model will be used by the visual and auditory object recognition groups. It has been checked into the learn group's svn repository. The repository is located at On a linux workstation it can be checked out into a user's home directory like so:

svn co ~/learn

Within that directory the LVis model is located in sims/lvis.proj.

The LVis model has been designed around the CU3D dataset located here: . It has a simple metadata format an example of which can be found at that URL under the section "Sample OID file".

The exact dataset that you want is:

In the subversion repository for the learn group unpack this directory inside the images directory. It should be called CU3D_100_renders_lr20_u30_nb.

You will also need the following metadata file inside the images/CU3D_100_renders_lr20_u30_nb directory:

Running LVis.proj

  • Open learn/sims/lvis.proj in emergent
  • Click Build when asked
  • Visit the middle panel tab called ControlPanel.
  • Click Init
  • Click cycle

Brian TODO

  • Create a testing config
  • Show others how to create metadata file
  • Robot camera taking images should be first thing we do with Nao
  • Upload the LVis paper to the wiki using the exclamation syntax to hide it rom the public.