Set of tools to access bonirob datasets in Python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
examples
pybonirob
tests
.gitignore
.gitlab-ci.yml
LICENSE.txt
Readme.md
setup.py

Readme.md

pybonirob

This package provides a basic set of tools to work with the bonirob datasets. It parses through the dataset and provides methods to access the data captured from different sensors. The dataset consists of measurements from a camera, kinect, velodyne, range scanner, gps and odometry.

The complete dataset can be downloaded from:

[http://www.ipb.uni-bonn.de/data/sugarbeets2016/]

Installation

Prerequisites

  • Requires python/python3: sudo apt-get install python python3
  • libpng, libfreetype : sudo apt-get install libpng-dev libfreetype6-dev

Module Dependencies

  • numpy
  • pyyaml
  • pillow
  • pytest (for running tests)
  • cv2 (for vizualization in the example)

These modules can be installed with

sudo pip install package-name

Install

To install the package:

cd /path/to/pybonirob
python setup.py install 

Usage

To load the dataset, we need to specify the location of the dataset to be loaded and then access different sensor measurements using the provided public methods.

Here is a basic example :

import pybonirob

base_path = '/path/to/the/dataset/directory'
prefix = 'bonirob_2016-05-23-10-47-22'
seq = '2'

# Data from different sensor modalities can be loaded separately
data.load_extrinsics()
data.load_camera()
data.load_gps()
data.load_laser()
data.load_odometry()

In the examples folder, a detailed demo program is provided which shows how to access the data from different sensors.

Related publication

If you use the dataset for your research, please cite the related publication:

@article{chebrolu2017ijrr,
title = {Agricultural robot dataset for plant classification, localization and mapping on sugar beet fields},
author = {Nived Chebrolu and Philipp Lottes and Alexander Schaefer and Wera Winterhalter and Wolfram Burgard and Cyrill Stachniss},
journal = {The International Journal of Robotics Research},
year = {2017}
doi = {10.1177/0278364917720510},
}