Source: own illustration based on images by DB Netz AG
Project Info
The "Open Sensor Data for Rail 2023" (OSDaR23, 10.57806/9mv146r0) has been created in a joint research project by the German Centre for Rail Traffic Research at the Federal Railway Authority (DZSF), Digitale Schiene Deutschland / DB Netz AG, and FusionSystems GmbH. Research report and Labeling Guide can be obtained from the DZSF website.
The data set consists of 45 sequences of annotated multi-sensor data (color camera, infrared camera, lidar, radar, localization, IMU). Data have been collected on different railway tracks in Hamburg, Germany.
License Info
The "Open Sensor Data for Rail 2023" (OSDaR23, 10.57806/9mv146r0) is published by the German Centre for Rail Traffic Research at the Federal Railway Authority (DZSF). Annotation data (file type .json
) are published under CC0 1.0. Sensor data (file types .png
, .pcd
, and .csv
) are published under CC BY-SA 3.0 de.
Further Info
The data set can be used in Python with the RailLabel package published by DB Netz AG.
The data set can be viewed, for example, with the WebLabel Player published by Vicomtech Research Foundation. (Disclaimer: Vicomtech was not part of the research project and there are currently no further relationships between DZSF and Vicomtech.)
Statistics
Number of Multisensor-Frames: 1534
Statistic of annotation objects:
object class number of annotations
------------------ ---------------------
person 73 421
crowd 1 352
train 8 290
wagons 110
bicycle 1 779
group of bicycles 644
motorcycle 14
road vehicle 12 669
animal 3 288
group of animals 0
wheelchair 0
drag shoe 79
track 18 543
transition 636
switch 2 947
catenary pole 27 706
signal pole 14 374
signal 32 790
signal bridge 312
buffer stop 4 539
flame 410
smoke 188
------------------ ---------------------
total 204 091