Hello, are there details on how well ROS packages tod_training and tod_detecting performed at "Solutions in Perceptions Challenge 2011" at ICRA in Shanghai 2011? These two ROS packages have been used for comparison to solutions submitted by participating contesters. According to the score function, winner's team got ca. 69% of max. score and tod_training+tod_detecting combo got 50.86%. The results are nice, but I am wondering about the tod_training+tod_detecting combo:
- choice of algorithms
- choice of hyperparameters
- used svn revision
- handling of limited computation times
For those accustomed to tod_*, it burns down to the question, whether one of the svn revisions has been used in the contest without further modification, and what the features.config.yaml and config.yaml (describing the parameter set) have looked like.
btw, this is the report which I am missing as well as the parameter configuration.
Originally posted by Julius on ROS Answers with karma: 960 on 2011-05-20
Post score: 2
Original comments
Comment by Julius on 2011-05-29:
I also attempted to ask guys from the challenge directly. Maybe some of the information has been presented only verbally at ICRA itself. There's no presentations or details on the Challenge website (neither on YouTube). Anyone in the community had the opportunity to watch the Challenge in Shanghai?
Comment by Julius on 2011-05-29:
I have also been foraying for further information on the OpenCV mailing list, see http://tech.groups.yahoo.com/group/OpenCV/message/80424 . Anybody investigating into this matter as well?