Skip to content
Go to file


Failed to load latest commit information.
Latest commit message
Commit time


KittiClass performance road type classification using Kitti Data. The model is able to distinguish between different road types. The data was generated by combining GPS information with Open Street Map information. Check out the paper of Wei-Chiu .et al for a desciption on how the data is generated. Out MultiNet comes with a detailed desciption of this model.

This project is very similar to the KittiSeg and KittiBox project. Unfortunately the training data for KittiClass is not public. You can still use to view results or use to train on your own data. Latter will however not work out of the box. If you are new to tensorflow, I recommend to use KittiSeg as starting point. The code is build to be compatible with the TensorVision backend which allows to organize experiments in a very clean way.


The code requires Tensorflow 1.0 as well as the following python libraries:

  • matplotlib
  • numpy
  • Pillow
  • scipy

Those modules can be installed using: pip install numpy scipy pillow matplotlib or pip install -r requirements.txt.


  1. Clone this repository:
  2. Initialize all submodules: git submodule update --init --recursive


Getting started

Run: python --input_image data/demo/demo.png to obtain a prediction using demo.png as input.

Run: python --hypes hypes/KittiClass.json train a KittiClass model.

Manage Data Storage

KittiSeg allows to separate data storage from code. This is very useful in many server environments. By default, the data is stored in the folder KittiClass/DATA and the output of runs in KittiClass/RUNS. This behaviour can be changed by setting the bash environment variables: $TV_DIR_DATA and $TV_DIR_RUNS.

Include export TV_DIR_DATA="/MY/LARGE/HDD/DATA" in your .profile and the all data will be downloaded to /MY/LARGE/HDD/DATA/data_road. Include export TV_DIR_RUNS="/MY/LARGE/HDD/RUNS" in your .profile and all runs will be saved to /MY/LARGE/HDD/RUNS/KittiClass

Modifying Model & Train on your own data

The model is controlled by the file hypes/KittiClass.json. Modifying this file should be enough to train the model on your own data and adjust the architecture according to your needs. A small sample of the train and val is provided.

Once again, I would like to point you to KittiSeg documentation. It contains more explanation on how to utilize your own data.


If you benefit from this code, please consider citing our paper:

  title={MultiNet: Real-time Joint Semantic Reasoning for Autonomous Driving},
  author={Teichmann, Marvin and Weber, Michael and Zoellner, Marius and Cipolla, Roberto and Urtasun, Raquel},
  journal={arXiv preprint arXiv:1612.07695},

If you use the Classification Data please cite:

  title={Find your Way by Observing the Sun and Other Semantic Cues},
  author={Ma, Wei-Chiu and Wang, Shenlong and Brubaker, Marcus A and Fidler, Sanja and Urtasun, Raquel},
  journal={arXiv preprint arXiv:1606.07415},


No releases published


No packages published