(Work in Progress) Very early stage development that I do in my free time now. This might break once in a while.
SuperSLAM is a deep learning based visual SLAM system that combines recent advances in learned feature detection and matching with the mapping capabilities of ORB_SLAM2.
It utilizes SuperPoint for keypoint detection and description and SuperGlue for robust feature matching between frames. These matches are then used by ORB_SLAM2 to estimate camera poses and build a map of the environment.
One of the future goals is to eventually replace the ORB_SLAM2 backend with a more modern and performant implementation and also release the code under a MIT Licence.
- CUDA==11.6
- TensorRT==8.4.1.5
- OpenCV>=4.0
- Eigen
- yaml-cpp
- DBoW3
- DBoW2
- Ubuntu 20.04
Clone the repository and the submodules.
git clone https://github.com/adityamwagh/SuperSLAM.git --recursive
cd SuperSLAMsh ./build_dependencies.shYou can use the included script to build the dependencies or install using the APT package manager.
# OpenCV
sudo apt-get install -y libopencv-dev
# Eigen
sudo apt install libeigen3-dev
# Pangolin
git clone --recursive https://github.com/stevenlovegrove/Pangolin.git
# Pangolin is split into a few components so you can include just what you need. Most dependencies are optional so you can pick and mix for your needs.
# Rather than enforcing a particular package manager, you can use a simple script to generate a list of (required, recommended or all) packages for installation for that manager (e.g. apt, port, brew, dnf, pacman, vcpkg):
# See what package manager and packages are recommended
./scripts/install_prerequisites.sh --dry-run recommended
# Override the package manager choice and install all packages
./scripts/install_prerequisites.sh -m brew allThe converted model is already provided in the weights folder, if you are using the pretrained model officially provided by SuperPoint and SuperGlue, you do not need to go through this step.
python convert2onnx/convert_superpoint_to_onnx.py --weight_file superpoint_pth_file_path --output_dir superpoint_onnx_file_dir
python convert2onnx/convert_superglue_to_onnx.py --weight_file superglue_pth_file_path --output_dir superglue_onnx_file_dirgit clone https://github.com/adityamwagh/SuperSLAM.git
cd SuperSLAM
sh ./build.shThe default image size param is 320x240, if you need to modify the image size in the config file, you should delete the old .engine file in the weights dir.
Lot of code is borrowed from these repositories. Thanks to the authors for opensourcing them!