https://github.com/klintan/ros2_pytorch

ROS2 PyTorch template node - running PyTorch C++ models in ROS2
https://github.com/klintan/ros2_pytorch

Last synced: 3 months ago
JSON representation

ROS2 PyTorch template node - running PyTorch C++ models in ROS2

• Host: GitHub
• URL: https://github.com/klintan/ros2_pytorch
• Owner: klintan
• License: apache-2.0
• Created: 2019-04-09T07:28:56.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-04-14T05:12:11.000Z (almost 6 years ago)
• Last Synced: 2024-08-05T00:04:21.211Z (6 months ago)
• Language: C++
• Size: 40.4 MB
• Stars: 39
• Watchers: 3
• Forks: 6
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# ROS2 PyTorch node

Run a PyTorch model in C++ in ROS2. A template node to be able to adapt to whatever model you want to run, for instance object detection of people and bikes.

Features

- 

There might be major changes to the code as it is a WIP.

### Topics

`/` - topic for 

## Build and installation

Make sure to run setup.bash and local_setup.bash for your ROS2 environment for all dependencies or symlink them into the repo.

Add Libtorch to your `CMAKE_PREFIX_PATH` env variable, like so:

```

export CMAKE_PREFIX_PATH=$CMAKE_PREFIX_PATH:/path/to/libtorch

```

Add CV_Bridge as well if its not already on your `CMAKE_PREFIX_PATH`

`export CMAKE_PREFIX_PATH=$CMAKE_PREFIX_PATH:` 

(If its built using `colcon build` this is the path to `cv_bridge` folder in the `install` folder.)

Run

`colcon build` 

Unfortunately you also have to copy two libs from https://github.com/intel/mkl-dnn/releases into the libtorch lib folder.

`libmklml.dylib`

`libiomp5.dylib`

## Usage

`ros2 run ros2_pytorch ros2_pytorch_node`

## References