DrivingGaussian: Composite Gaussian Splatting for Surrounding Dynamic Autonomous Driving Scenes
This work addresses the challenge of accurately reconstructing complex, dynamic driving scenes for autonomous vehicle applications, representing an incremental advancement over existing methods.
The paper tackles the problem of reconstructing dynamic autonomous driving scenes with moving objects by proposing DrivingGaussian, which uses composite Gaussian splatting to model static backgrounds and dynamic objects, resulting in improved reconstruction and photorealistic surround-view synthesis with high-fidelity and multi-camera consistency.
We present DrivingGaussian, an efficient and effective framework for surrounding dynamic autonomous driving scenes. For complex scenes with moving objects, we first sequentially and progressively model the static background of the entire scene with incremental static 3D Gaussians. We then leverage a composite dynamic Gaussian graph to handle multiple moving objects, individually reconstructing each object and restoring their accurate positions and occlusion relationships within the scene. We further use a LiDAR prior for Gaussian Splatting to reconstruct scenes with greater details and maintain panoramic consistency. DrivingGaussian outperforms existing methods in dynamic driving scene reconstruction and enables photorealistic surround-view synthesis with high-fidelity and multi-camera consistency. Our project page is at: https://github.com/VDIGPKU/DrivingGaussian.