ROS-packages

A collection of ROS packages for LMB software

License

Non-commercial, for research purposes only.

Contents

Most packages in this repository are neural networks based on the Caffe library. Use our modified legacy Caffe code.

Disparity

• LMB_dispnet runs DispNet disparity estimation networks -- mostly the newer versions.
• LMB_disparity_view converts the outputs of LMB_dispnet into displayable images.

Optical flow

• LMB_flownet and LMB_flow_view do the same thing as LMB_dispnet and LMB_disparity_view (see above), but for optical flow.

Scene flow

• LMB_sceneflow runs LMB_dispnet and LMB_flownet nodes for scene flow (= dense 3D motion vectors) estimation.
• LMB_sceneflow_vis contains various visualizations for scene flow.

Data input

• LMB_generic_images_feeder is a simple node that reads loose collections of image files and produces ROS image topics.

Installation

• Get additional binary package with pretrained network weights from LMB
• Build using catkin (tested with ROS Kinetic)

Author: Nikolaus Mayer, 2019

This repository does not accept pull requests.

The software in this repository is part of the Horizon2020 project Trimbot2020.