Sparse-Dense Motion Modelling and Tracking for Manipulation without Prior Object Models

This repo contains the RGB-D sequences used for the paper "Sparse-Dense Motion Modelling and Tracking for Manipulation without Prior Object Models" in the ROS 1 bag format.

The bag files contain the colour images with intrinsics, the depth images registered to the colour frame and ground truth camera poses:

/depth_to_rgb/camera_info                    : sensor_msgs/CameraInfo
/depth_to_rgb/image_raw/compressed           : sensor_msgs/CompressedImage
/depth_to_rgb/image_raw/filtered/compressed  : sensor_msgs/CompressedImage
/diagnostics                                 : diagnostic_msgs/DiagnosticArray
/rgb/camera_info                             : sensor_msgs/CameraInfo
/rgb/image_raw/compressed                    : sensor_msgs/CompressedImage
/tf                                          : tf2_msgs/TFMessage
/tf_static                                   : tf2_msgs/TFMessage
/vicon/kinect/kinect                         : geometry_msgs/TransformStamped
/vicon/markers                               : vicon_bridge/Markers

If you use this data, please cite:

@ARTICLE{Rauch2022,
  author={Rauch, Christian and Long, Ran and Ivan, Vladimir and Vijayakumar, Sethu},
  journal={IEEE Robotics and Automation Letters},
  title={Sparse-Dense Motion Modelling and Tracking for Manipulation Without Prior Object Models},
  year={2022},
  volume={7},
  number={4},
  pages={11394-11401},
  doi={10.1109/LRA.2022.3200177}
}

Mirrors:

Usage

ROS 1

Use the ROS 1 rosbag command-line tool to publish the topics from the bag files:

rosbag play estimation/nx_estim1_manipulation.bag

ROS 2

Install rosbags:

pip3 install rosbags

and convert the bag files from ROS 1 to ROS 2:

find . -type f -name *.bag -exec rosbags-convert "{}" \;

Then you can use the ROS 2 command-line tool to publish the topics:

ros2 bag play estimation/nx_estim1_manipulation

About

https://conferences.inf.ed.ac.uk/MultiMotionFusion/