pyORBSLAM2
Ultra-fast Boost.Python interface for ORBSLAM2
Are you using ORBSLAM2 for your latest tracking project? Do you just want to create a quick prototype, or are you weary of writing C++ code? Then this repository is for you! Simply interface ORBSLAM2 from python2.x/3, using your favourite NumPy arrays as necessary. It is as easy as:
import ORBSLAM2 as os2
slam_obj = os2.SLAM()
slam_obj.init("/slamdoom/libs/orbslam2/Vocabulary/ORBvoc.txt",\
"/slamdoom/libs/orbslam2/Examples/RGB-D/TUM1.yaml")
slam_obj.track(array, array_d, time.time())
del slam_obj
You can use this repo as a basis for any other Boost.Python projects as well.
Please contact me at cs@robots.ox.ac.uk any time and I will assist you!
Best, Christian Schroeder de Witt, Torr Vision Group, University of Oxford, 2017
Get started
Use run.sh to run nvidia-docker environment to build build/ORBSLAM2.so, which you should then put on your PYTHONPATH. See test/test.py for an example. Extend orbslam2.cpp to your needs (comes with NumPy/cv::Mat conversion routines, and already includes Boost::Python::NumPy set up for you should you need it)
See below more detailed steps on how to set up the tutorial examples.
Ubuntu
Clone this repository (or even better: fork it!)
git clone git@github.com:torrvision/pyORBSLAM2.git
Install nvidia-docker plugin: see https://github.com/NVIDIA/nvidia-docker Note: You require nvidia-docker plugin on host, as Pangolin2 requires OpenGL support
Now run the nvidia-docker container:
sudo ./run.sh
(It will take a while to compile OpenCV3) (SSH password: source)
Now, inside the container SSH shell, simply do: (ensure build and bin folders are deleted to force CMake to update properly)
/orbslam/src/build.sh
export PYTHONPATH=/orbslam/src/build:$PYTHONPATH
Now you should be able to try the test example inside the container:
python3 /orbslam/src/test/test.py
(press key for GUI windows to open)
The above is for Python3 only, if you require Python2.x support or have any other questions / issues, please contact me:
All the best
Christian