This is the official repository for the CVPR 2025 paper SplatAD: Real-Time Lidar and Camera Rendering with 3D Gaussian Splatting for Autonomous Driving. The code in this repository builds upon the open-source library gsplat, with modifications and extensions designed for autonomous driving data.
While the code contians all components needed to efficiently render camera and lidar data, the SplatAD-model itself, including dataloading, decoders, etc., will be released through neurad-studio.
We welcome all contributions!
- Efficient lidar rendering
- Projection to spherical coordinates
- Depth and feature rasterization for a non-linear grid of points
- Rolling shutter compensation for camera and lidar
Our code introduce no additional dependencies. We thus refer to the original documentation from gsplat for both installation and development setup.
See rasterization and lidar_rasterization for entry points to camera and lidar rasterization.
Additionally, we provide example notebooks under examples that demonstrate lidar rendering and rolling shutter compensation.
For further examples, check out the test files.
- gsplat - Collaboration friendly library for CUDA accelerated rasterization of Gaussians with python bindings
- 3dgs-deblur - Inspiration for the rolling shutter compensation
You can find our paper on arXiv.
If you use this code or find our paper useful, please consider citing:
@article{hess2024splatad,
title={SplatAD: Real-Time Lidar and Camera Rendering with 3D Gaussian Splatting for Autonomous Driving},
author={Hess, Georg and Lindstr{\"o}m, Carl and Fatemi, Maryam and Petersson, Christoffer and Svensson, Lennart},
journal={arXiv preprint arXiv:2411.16816},
year={2024}
}
