Toward Real-world BEV Perception: Depth Uncertainty Estimation via Gaussian Splatting
This work improves BEV perception for autonomous driving systems by enhancing uncertainty modeling and efficiency, though it is incremental as it builds on the Lift-Splat-Shoot paradigm.
The paper tackles the problem of Bird's-eye view (BEV) perception for autonomous driving by addressing the lack of uncertainty modeling and high computational costs in existing methods, introducing GaussianLSS which achieves state-of-the-art performance on the nuScenes dataset with 2.5x faster speed and 0.3x less memory usage while maintaining competitive accuracy.
Bird's-eye view (BEV) perception has gained significant attention because it provides a unified representation to fuse multiple view images and enables a wide range of down-stream autonomous driving tasks, such as forecasting and planning. Recent state-of-the-art models utilize projection-based methods which formulate BEV perception as query learning to bypass explicit depth estimation. While we observe promising advancements in this paradigm, they still fall short of real-world applications because of the lack of uncertainty modeling and expensive computational requirement. In this work, we introduce GaussianLSS, a novel uncertainty-aware BEV perception framework that revisits unprojection-based methods, specifically the Lift-Splat-Shoot (LSS) paradigm, and enhances them with depth un-certainty modeling. GaussianLSS represents spatial dispersion by learning a soft depth mean and computing the variance of the depth distribution, which implicitly captures object extents. We then transform the depth distribution into 3D Gaussians and rasterize them to construct uncertainty-aware BEV features. We evaluate GaussianLSS on the nuScenes dataset, achieving state-of-the-art performance compared to unprojection-based methods. In particular, it provides significant advantages in speed, running 2.5x faster, and in memory efficiency, using 0.3x less memory compared to projection-based methods, while achieving competitive performance with only a 0.4% IoU difference.