Skip to content

Latest commit

 

History

History
43 lines (30 loc) · 2.2 KB

README.md

File metadata and controls

43 lines (30 loc) · 2.2 KB

rosbag2csv

This program converts a rosbag file into a csv file, where the individual messages are broken down into scalar values. Images are saved in a separate folder and linked to from the csv file via their name. Users can specify topics to extract, the intervals between extraction, the starting point and timeframe of extraction, etc.

Installation

Users must have ROS installed (f.e. Noetic). Additionally, Numpy and cv2 must be installed (f.e. via pip).

Usage

The --help explains everything:

usage: python3 rosbag2csv.py [-h] [-o] [-d] [-l] input_filepath interval topics [topics ...]

Extracts data from a .bag file into a csv with frequency synchronization and image extraction

positional arguments:
  input_filepath   The path to the input rosbag-file. Can be relative or absolute.
  interval         The interval with which extraction out of the rosbag file takes place. Should not be smaller than
                   1/lowest_frequency.
  topics           The topics that should be processed on the rosbag file. Can be supplied via multiple command line
                   arguments, a json file (a single list), or a mixture of both. Write "all" to include all topics.

optional arguments:
  -h, --help       show this help message and exit
  -o , --out-dir   The path to the output directory. (Default: the current
                   directory)
  -d , --delay     The number of seconds from the beginning of the rosbag-file that the extraction should be delayed by.
                   (Default: 0)
  -l , --length    The length of the extraction period in seconds. (Default: Until end of rosbag-file)

NOTE: Don't use topics that use the nav_msgs/Path ROS message type. The first save takes place at <start_time> + <interval_time>. The last (partially finished) interval are excluded. An empty output file could mean that you specified the topics wrongly

Acknowledgements

This tool was originally desgined for use on the terrasentia dataset of Jose Cuaran (github.com/jrcuaranv/terrasentia-dataset). The extraction of images was adapted from his repository.

Written by Thomas Woehrle (github.com/thomas-woehrle) in January 2024 at the University of Illinois Urbana-Champaign. Supervised by Arun V Sivakumar.