{"name":"GigaSLAM: Large-Scale Monocular SLAM with Hierarchical Gaussian Splats","description":"GigaSLAM is a groundbreaking monocular SLAM framework designed for kilometer-scale outdoor environments. It leverages hierarchical Gaussian splats and neural networks to achieve efficient, scalable mapping and high-fidelity rendering. This system addresses the challenges of large-scale tracking and mapping using only RGB input, extending the applicability of Gaussian Splatting SLAM to unbounded outdoor scenes.","github":"https://github.com/DengKaiCQ/GigaSLAM","url":"https://osrepos.com/repo/dengkaicq-gigaslam","source":"osrepos.com","sourceDescription":"This repository profile is provided by osrepos.com, an open source repository discovery platform.","repositoryProfile":"https://osrepos.com/repo/dengkaicq-gigaslam","generatedFor":"open source discovery and AI-assisted research","markdown":"https://osrepos.com/repo/dengkaicq-gigaslam.md","json":"https://osrepos.com/repo/dengkaicq-gigaslam.json","topics":["SLAM","Monocular SLAM","Gaussian Splatting","NeRF","Computer Vision","Robotics","C++","Deep Learning"],"keywords":["SLAM","Monocular SLAM","Gaussian Splatting","NeRF","Computer Vision","Robotics","C++","Deep Learning"],"stars":null,"summary":"GigaSLAM is a groundbreaking monocular SLAM framework designed for kilometer-scale outdoor environments. It leverages hierarchical Gaussian splats and neural networks to achieve efficient, scalable mapping and high-fidelity rendering. This system addresses the challenges of large-scale tracking and mapping using only RGB input, extending the applicability of Gaussian Splatting SLAM to unbounded outdoor scenes.","content":"## Introduction\n\nGigaSLAM introduces the first RGB NeRF / 3D Gaussian Splatting (3DGS)-based SLAM framework specifically engineered for kilometer-scale outdoor environments. Traditional Neural Radiance Fields (NeRF) and 3DGS SLAM methods are typically confined to smaller, bounded indoor settings. GigaSLAM overcomes these limitations by employing a hierarchical sparse voxel map representation, where Gaussians are decoded by neural networks at multiple levels of detail. This innovative design enables efficient, scalable mapping and high-fidelity viewpoint rendering across expansive, unbounded scenes, as demonstrated on challenging datasets like KITTI, KITTI 360, 4 Seasons, and A2D2.\n\n## Installation\n\nTo get GigaSLAM up and running, follow these steps. Note that the project relies on CUDA/C++ components and benefits from a robust hardware environment, ideally with 32+ GiB CPU RAM for processing ultra-long sequences.\n\n### 1. Hardware and System Environment\n\nThe project was developed and tested on systems with NVIDIA RTX 4090/L20 GPUs, Intel Xeon CPUs, and Ubuntu 22.04.3 LTS with CUDA 11.8.\n\n### 2. Environment Setup\n\nFirst, create a conda virtual environment and install PyTorch (version 2.2.0 with CUDA 11.8 is recommended):\n\nbash\nconda create -n gigaslam python=3.10\nconda activate gigaslam\npip install torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu118\npip install -r requirements.txt\n\n\nSpecial attention should be paid to `torch_scatter` and `xformers` dependencies. If `torch_scatter` installation fails, manually download and install the wheel from the PyTorch Geometric website. For `xformers-0.0.24`, ensure compatibility with `PyTorch 2.2.0` to avoid forced uninstallation of your PyTorch version.\n\n### 3. Compiling CUDA/C++ Modules\n\nInstall the necessary CUDA/C++ components:\n\n*   **3D GS Rendering Module:**\n    bash\npip install submodules/simple-knn\npip install submodules/diff-gaussian-rasterization\n    \n*   **Loop-Closure Detection Module:** Install `libopencv-dev`, compile and install `DBoW2`, then install `DPRetrieval`.\n    bash\nsudo apt-get install -y libopencv-dev\ncd DBoW2\nmkdir -p build && cd build\ncmake ..\nmake\nsudo make install\ncd ../..\npip install ./DPRetrieval\n    \n*   **Loop-Closure Correction Module:**\n    bash\npython setup.py install\n    \n\n### 4. Bag of Words Model Setup\n\nDownload and extract the pre-trained Bag of Words vocabulary for `DBoW2`:\n\nbash\nwget https://github.com/UZ-SLAMLab/ORB_SLAM3/raw/master/Vocabulary/ORBvoc.txt.tar.gz\ntar -xzvf ORBvoc.txt.tar.gz\n\n\n## Examples\n\nTo run GigaSLAM, you first need to modify the `.yaml` configuration files located in the `./config` directory. Update the `Dataset` path to your local dataset (in PNG or JPG format) and specify the camera intrinsics (`fx`, `fy`, `cx`, `cy`).\n\nFor instance, an example configuration might look like this:\n\nyaml\nDataset:\n  color_path: \"/media/deng/Data/4SeasonsDataset/BusinessCampus_recording_2020-10-08_09-30-57/undistorted_images/cam0\"\n  Calibration:\n    fx: 501.4757919305817\n\n\nOnce configured, execute the SLAM process using the following command:\n\nbash\npython slam.py --config ./configs/kitti_06.yaml\n\n\nPretrained weights for `DISK`, `LightGlue`, and `UniDepth` will be automatically downloaded on the first execution. If you encounter issues loading `UniDepth` models from HuggingFace due to network restrictions, you can try setting `export HF_ENDPOINT=https://hf-mirror.com` or manually download the model weights and specify their local path in the configuration.\n\nIf `['SLAM']['viz']` is set to `True` in your `.yaml` file, you will observe real-time visualization outputs in the `results/your_exps/` directory during execution, showcasing the tracking and mapping process.\n\n## Why Use GigaSLAM?\n\nGigaSLAM offers significant advantages for researchers and developers working in computer vision and robotics:\n\n*   **Unprecedented Scale**: It is the first RGB NeRF/3DGS-based SLAM system capable of operating in kilometer-scale outdoor environments, a major breakthrough for this class of methods.\n*   **High-Fidelity Mapping**: Utilizes hierarchical Gaussian splats and neural networks to achieve highly detailed and visually faithful 3D reconstructions.\n*   **Robust Tracking**: Employs a sophisticated front-end tracking mechanism combining a metric depth model, epipolar geometry, PnP algorithms, and a Bag-of-Words-based loop closure for accurate and consistent pose estimation over long trajectories.\n*   **Efficiency and Scalability**: The hierarchical sparse voxel map representation ensures efficient data handling and scalability, making it suitable for large and unbounded scenes.\n*   **Monocular Input**: Achieves impressive results using only monocular RGB input, reducing sensor complexity and cost.\n\n## Links\n\n*   GitHub Repository: [GigaSLAM GitHub](https://github.com/DengKaiCQ/GigaSLAM){:target=\"_blank\"}\n*   arXiv Paper: [GigaSLAM: Large-Scale Monocular SLAM with Hierarchical Gaussian Splats](https://arxiv.org/abs/2503.08071){:target=\"_blank\"}","metrics":{"detailViews":8,"githubClicks":12},"dates":{"published":null,"modified":"2026-02-28T16:52:31.000Z"}}