GaussianFusionOcc: A Seamless Sensor Fusion Approach for 3D Occupancy Prediction Using 3D Gaussians
This addresses the problem of precise environment interpretation for autonomous driving systems, representing an incremental advancement through novel fusion and representation techniques.
The paper tackles 3D semantic occupancy prediction for autonomous driving by proposing GaussianFusionOcc, which uses semantic 3D Gaussians and a sensor fusion mechanism to integrate camera, LiDAR, and radar data, resulting in improved memory efficiency, inference speed, and outperforming state-of-the-art models.
3D semantic occupancy prediction is one of the crucial tasks of autonomous driving. It enables precise and safe interpretation and navigation in complex environments. Reliable predictions rely on effective sensor fusion, as different modalities can contain complementary information. Unlike conventional methods that depend on dense grid representations, our approach, GaussianFusionOcc, uses semantic 3D Gaussians alongside an innovative sensor fusion mechanism. Seamless integration of data from camera, LiDAR, and radar sensors enables more precise and scalable occupancy prediction, while 3D Gaussian representation significantly improves memory efficiency and inference speed. GaussianFusionOcc employs modality-agnostic deformable attention to extract essential features from each sensor type, which are then used to refine Gaussian properties, resulting in a more accurate representation of the environment. Extensive testing with various sensor combinations demonstrates the versatility of our approach. By leveraging the robustness of multi-modal fusion and the efficiency of Gaussian representation, GaussianFusionOcc outperforms current state-of-the-art models.