Start Date:
End Date:

Introduction of the new KITTI Importer component from RTMaps which provides an intuitive interface for integrating the KITTI dataset for Robotics and ADAS applications.



NEW RTMaps Package

KITTI Sensor Datasets Importer

The KIT (Karlsruhe Institute of Technology Fig.1) was created in 1954, and is the result of fusion of the University of Karlsruhe and the Center of Research of Karlsruhe.

 Karlsruhe Institute of Technology
Fig. 1: Karlsruhe Institute of Technology

The KITTI Dataset is a free and publicly-available collection of recordings captured from a variety of sensors.

It has been produced from a moving platform while driving in and around Karlsruhe in Germany.
The platform used is a VW Passat B6

KITTI Car Setup

  • 4 Point Grey FLEA 2 Cameras
  • 1 LIDAR Velodyne HDL-64E
  • 1 GPS/IMU unit from OXTS
Sensor Setup from the KITTI Vision Benchmark Suite
Fig. 2: Sensor Setup from the KITTI Vision Benchmark Suite

For more information you can see this official presentation video for The KITTI Vision Benchmark Suite by Andreas Geiger

The captured data have been sorted in several categories (City, Residential, Road, Campus ...) and a significant portion of them have been synchronized, adjusted and made available on the KITTI website as zip files (Fig. 3)

typical structure of a KITTI zip file
Fig. 3: The typical structure of a KITTI zip file (crédit from ) 


KITTI Importer

In order to facilitate the use of the dataset in RTMaps, we have developed the KITTI Importer component (Fig. 4) which provides an intuitive interface for integrating the KITTI dataset in Vision, Robotics and ADAS applications.

Fig.5 KITTI importer
Fig. 4: KITTI Importer

The component reads synchronized + rectified data from a KITTI dataset and makes them available on its outputs.
It is also possible to read the ‘ground truth’ data (i.e. tracklets) from the tracklet_labels.xml file when available.

In addition to taking care of the data parsing, KITTI Importer allows 2 Replay Modes:

  • Immediate: The data are output as fast as possible,
  • Clocked: The data are synchronized in concordance with the timestamps of the data set (typically at 10Hz).

For optimal performance at playback and random access capabilities, we recommend to play and re-record the KITTI data in the RTMaps data format, (using the Recorder component). This file format handles index files which allow instantaneous random-access capabilities to the dataset and the RTMaps Player component will provide multithreaded playback and early data extraction for achieving optimal performance.

To illustrate how straightforward it is to read the KITTI dataset using our new importer, we have made the following video demonstration. It walks you through the placement, the configuration and the visualization of the data in the RTMaps Studio interface.

More informations on Kitti :

Official Paper :
Kitti Setup Document :
Datasets Kitti Raw Data :
Calculations (Using the DevKit) :