tesse-segmentation-ros

A package for running semantic segmentation networks in ROS using TensorRT.

Usage

To Run

To run the segmentation node in TESSE, use the provided launch file:

Note: The required model definition file is not included in this repository. To run a trained network, download one of the provided ONNX files from the the release page and place it in the ./cfg folder. Then, pass the appropriate path to the weight_file argument in the provided launch file:

roslaunch tesse_segmentation_ros tesse_segmentation_ros.launch weight_file:=PATH_TO_WEIGHT_FILE

Training a new network

See training to get started on training a network in TESSE.

Installation

TensorRT

This requires TensorRT 6.0.0. To install via a tar file

• Download a TensorRT tar file compatible with your machine through the NVIDIA TensorRT page
• Unpack the tar file into a directory, <TRT_DIR>
• Add the absolute path to the TensorRT lib directory to your LD_LIBRARY_PATH

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:<TRT_DIR>/lib

• Install the Python wheel file matching your Python version.

cd <TRT_DIR>/python
sudo pip install tensorrt-<TRT_VERSION>-cp<PYTHON_VERSION>-none-linux_x86_64.whl

Install PyCUDA

pip install pycuda

Clone and build this repo

# setup catkin workspace
mkdir -p ~/catkin_ws/src
cd ~/catkin_ws
catkin init

cd src 
git clone git@github.com:MIT-TESSE/tesse-segmentation-ros.git
cd ..

# install dependencies
wstool init
wstool merge tesse-segmentation-ros/install/tesse_segmentation_ros.rosinstall 
cd ..

# compile
catkin build

# source workspace
source ~/catkin_ws/devel/setup.bash

Disclaimer

DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited.

This material is based upon work supported by the Under Secretary of Defense for Research and Engineering under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Under Secretary of Defense for Research and Engineering.

(c) 2020 Massachusetts Institute of Technology.

MIT Proprietary, Subject to FAR52.227-11 Patent Rights - Ownership by the contractor (May 2014)

The software/firmware is provided to you on an As-Is basis

Delivered to the U.S. Government with Unlimited Rights, as defined in DFARS Part 252.227-7013 or 7014 (Feb 2014). Notwithstanding any copyright notice, U.S. Government rights in this work are defined by DFARS 252.227-7013 or DFARS 252.227-7014 as detailed above. Use of this work other than as specifically authorized by the U.S. Government may violate any copyrights that exist in this work.