SRVC Meeting minutes for September 9th, 2009

• Laptop situation:
  • We planned some changes to the structure of our development environment to stop the rapidly growing requirement for laptops before it overwhelms Jim's budget.
  • We will now try to mount each of the 3-4 working lab laptops on the robot permanently and keep them maintained in proper working condition so that we always have a testing environment.
  • Marius will install Ubuntu on the newest laptop and we'll use that as the base machine. The other lab laptops will be added shortly afterwards. [MARIUS, DAVE]
  • Development should now be done on desktops, either by setting the desktop up with the required libraries etc or by sshing to the development laptops
  • For the contest, we may try to add Scott's mac for some extra computation

• Short discussion about moving the robot to Vegas. Catherine to investigate this over the week and report back next meeting [CATHERINE].

• Discussion about tilting laser
  • We'd like to decide on how to mount the tilting laser soon. This will potentially involve replacing the tripod and may change quite a bit in the system.
  • Investigate the options, come up with a solution and implement it. [MARIUS and CATHERINE]
  • All discussed the need for calibration between the laser and our other sensors. Consensus was this could be tricky and we'll just have to do the best we can. Revisit this later likely...

• Discussed how team members should get to Vegas. Luckily flights will be cheap and some team members can go with the robot if we decide to drive. Keep on top of this, but nothing active for now.

• Marius working on ROS conversion. Bumblebee is done as of this evening and next thing to look at is player on the robot. This will be ongoing for the next while.

• James presented his progress with the recognition harness.
  • He's implemented wrappers for training and detection. We discussed the format for features and methodology surrounding feature extraction and saving/loading features between classfiers. Scott and James will discuss soon.
  • James to commit the code soon and we should all have a look. [JAMES and ALL]

• Round table discussion of who's doing what:
  • Scott: working on integrating boundary recognizer with SIFT and stereo
  • Tomas: working on visual SLAM with Pooja
  • Catherine: working on the tilting laser and ROS integration
  • Ankur: evaluating recognizers on 3D viewpoint data (Felzenswalb currently)
  • Marius: ROS conversions, Ubuntu on the new laptop, did 3D model fitting stuff at WG which may be useful
  • Walter: using 3D warehouse models for object recognition. So far focused on spherical harmonics.
  • Dave: evaluating a variety of classifiers on the multiview dataset and implementing planning algorithms for the robot to collect good views
  • Pooyan: image filtering, web data gathering. Subsequent discussion branched into where to gather 3D models, and a variety of other datasources. Pooyan mentioned some notable changes in the final rules.
  • James: making the ROS recognition harness and leaving soon!

• TODO: Ask if mechanical turk is ok to use during SRVC training data gathering. [DAVE?]

-- DavidMeger - 10 Sep 2009