============================================================
== Getting Started 
============================================================


You'll need an Ubuntu machine with ROS C turtle installed.  (Other versions may work, but have not been tested.)  See http://www.ros.org/wiki/cturtle/Installation/Ubuntu.

Code and data are available at http://cs.stanford.edu/people/teichman/stc/.

The code is a ROS package and can be built with the command 'rosmake' from the code directory.  You'll need to add the code directory to the environment variable ROS_PACKAGE_PATH first.

Once the code builds, try out the example program, e.g.
bin/example ../<stc path>/natural/pass2-10-16-2009_15-00-46.tm


============================================================
== Details
============================================================

The data is split into 'natural' and 'background' directories.  Data in 'natural' are from normal street scenes and are hand labeled.  Data in 'background' are from street scenes that were known to have no pedestrian, bicyclists, or cars in them.

The training / testing split used in the ICRA2011 paper is as follows:

- Training data

lasuen02-11-17-2009_17-07-43
CampusLoop3-10-03-2009_15-32-37-FirstMinute
pass2-10-16-2009_15-00-46
lomita_and_santa_teresa01-11-17-2009_18-35-12
intersection-04-07-2010_09-05-08
intersection-04-07-2010_09-21-53
ungrad_area-10-06-2010_16-12-40-cut
All the logs in the background/ dir.

- Testing data

dish_area02-11-17-2009_20-13-49
page_mill_and_hansen01-11-17-2009_19-49-18
page_mill_and_hansen02-11-17-2009_19-56-57
white_plaza02-11-17-2009_17-35-42
driving_test_set-11-17-2009_19-13-35


Unfortunately, the intersection and pass2 datasets were collected without the localizer running.  If you are using the robot_pose estimates, watch out: they are wrong for these three logs.  Thanks to Thomas Schamm for bringing this to our attention.

============================================================
== Paper
============================================================

If you use this data in the course of your research, please cite the following paper:

@inproceedings{Teichman2011,
author = {Teichman, Alex and Levinson, Jesse and Thrun, Sebastian},
booktitle = {International Conference on Robotics and Automation},
title = {Towards 3D Object Recognition via Classification of Arbitrary Object Tracks},
year = {2011}
}


============================================================
== Contact
============================================================

Please see the image of my email address on the STC website:
http://cs.stanford.edu/people/teichman/stc/

Hopefully spammers haven't figured out OCR yet.


============================================================
== Updates
============================================================

1/6/2012
ungrad_area-10-06-2010_16-12-40-cut was mistakenly not included in the original tarball.  It has been added.  The md5sum of the new tarball is 73cfa8c9dc01b48bda064961371c2e13.