This project is a python wrapper to SLAM algorithms.
At the moment, we provide support for:
- ORB_SLAM2
- ORB_SLAM3
- the jupyter notebook example_usage shows how to use ORB_SLAM2 with a sequence of the KITTI dataset
- the jupyter notebook trajectory_example draw the camera trajectory and the point cloud
to change the algorithm settings you can modify the setting.yaml file in the line
SLAM.alg:'insert here your .py file that contains the wrapper class'
you can add at the file the other params that the method needs. to change from ORB_SLAM2 to ORB_SLAM3 you need to change only the SLAM.alg entry.
the params needed for this 2 algorithms are:
- a vocabulary file
- a setting file with the intrinsics parameters of the cameras and other configuration params
We provided the vocabulary file and the configuration file for ORB_SLAM2/3 for the KITTI_02 camera and the TUM freiburg3 camera in the slam_metohd/Settings folder, but you can add your owns using as model the files currently provided.
We provide a container on dockerhub, in which all the dipendences and the repository are already installed.
docker pull giordanolaminetti/slampy:tag
The tag correspond to the container you need:
basecorrespond to a container with installed only the dependencesorbslam2correspond to a container with only ORB_SLAM2 installedorbslam3correspond to a container with only ORB_SLAM2 installedlatestcorrespond to a container with ORB_SLAM2 and ORB_SLAM3 installed
When the image is ready, you can create a new container running:
NAME="orb"
DATAPATH="/PATH/TO/KITTI/DATE/DRIVE_SYNC_FOLDER/"
TAG="latest"
sudo docker run -it \
--name $NAME \
--mount type=bind,source="$(pwd)",target=/slampy/slampy \
-v $DATAPATH:"/slampy/slampy/Dataset":ro \
-p 8888:8888 --rm\
giordanolaminetti/slampy:$TAG /bin/bash
Doing so, the created container contains both the code and the Dataset (in read-only mode to prevent wrong behaviours)
You can have a test with the jupyter example running:
jupyter notebook --ip 0.0.0.0
Our project has been developed starting from other repositories, in particular:
- Python bindings to ORB Slam are based on the repo
- ORB Slam 2 has been cloned from the original repository
- ORB Slam 3 has been cloned from the original repository