ORB-SLAM3-PYTHON

Todo:

• when lost, don't create new map (have bug), might just assume not moving, or switch to optical flow based method (which is supper fast)

Python bindings generated using pybind11. We use a modified version of ORB-SLAM3 (included as a submodule) to exntend interfaces. It might not be the most up-to-date with the original ORB-SLAM3.

Update

• Oct. 3rd, 2023: Added demo code.
• Feb. 7th, 2023: First working version.

Dependancy

• OpenCV >= 4.4
• Pangolin
• Eigen >= 3.1
• C++11 or C++0x Compiler
• install octomap from source (latest version) https://github.com/OctoMap/octomap

Installation

1. Clone the repo with --recursive
2. Install Eigen, Pangolin and OpenCV if you havn't already.
3. ORB-SLAM3 requires openssl to be installed: sudo apt-get install libssl-dev
4. Run python setup install or pip install ..
5. Please raise an issue if you run into any.

Demo

Please see the demo at demo/run_rgb.py for how to use this code. For example, you can run this demo with (by substituting the appropriate arguments):

python demo/run_rgbd.py \
    --vocab_file=third_party/ORB_SLAM3/Vocabulary/ORBvoc.txt \
    --settings_file=third_party/ORB_SLAM3/Examples/RGB-D/TUM1.yaml \
    --dataset_path=/mnt/dataset2/TUM/rgbd_dataset_freiburg1_xyz