Tools for working with the KITTI dataset in Python.

The majority of this project is available under the MIT license. The files in kitti/bp are a notable exception, being a modified version of Pedro F. Felzenszwalb and Daniel P. Huttenlocher's belief propogation code 1 licensed under the GNU GPL v2. These files are not essential to any part of the rest of the project, and are only used to run the optional belief propogation disparity image interpolation.

To begin working with this project, clone the repository to your machine

git clone https://github.com/hunse/kitti

Download the KITTI data to a subfolder named data within this folder. Most of the tools in this project are for working with the raw KITTI data. For example, if you download and unpack drive 11 from 2011.09.26, it should be in the folder data/2011_09_26/2011_09_26_drive_0011_sync. The calibration files for that day should be in data/2011_09_26.

Since the project uses the location of the Python files to locate the data folder, the project must be installed in development mode so that it uses the original source folder

python setup.py develop

You should now be able to import the project in Python. If you have trouble with commands like kitti.raw.load_video, check that kitti.data.data_dir points to the correct location (the location where you put the data), and that commands like kitti.data.get_drive_dir return valid paths.

For examples of how to use the commands, look in kitti/tests. Most of the examples use drive 11, but it should be easy to modify them to use a drive of your choice.

Happy hacking!

The belief propagation module uses Cython to connect to the C++ BP code. To build the Cython module, run

python setup.py build_ext --inplace

This should create the file module.so in kitti/bp.