CRLab / pc_scene_completion_ws

ROS scene completion workspace and external dependencies.
http://shapecompletiongrasping.cs.columbia.edu/
16 stars 5 forks source link
deep-learning robotics shape-completion

Build Status

pc_scene_completion_ws

ROS scene completion workspace and external dependencies. This is a self contained workspace for running the pc_pipeline with the shape completion CNN. This is the workspace from the IROS 2017 paper "Shape Completion Enabled Robotic Grasping" found here:

http://shapecompletiongrasping.cs.columbia.edu/

and is demonstrated in the video below:

Shape Completion Enabled Robotic Grasping

Please read more about running and trouble shooting the pc_pipeline here:

https://github.com/CRLab/pc_pipeline_launch

Please read more about the CNN for shape completion here:

https://github.com/CRLab/pc_object_completion_cnn

Setup

git clone https://github.com/CRLab/pc_scene_completion_ws.git --recursive

Follow installation instructions found in the .travis.yml file. Specifically starting here:

https://github.com/CRLab/pc_scene_completion_ws/blob/master/.travis.yml#L50

You will need to install all the libraries in the external folder, and then you can build the src folder as a ros workspace. The travis.yml script does not cover Keras installation which is well documented on the Keras site. We run Keras with a tensorflow backend.