DrivingDepth: Sparse-Prompted Pixel-wise Scale Correction for Driving Depth Estimation
For autonomous driving, this method resolves the trade-off between metric accuracy and geometric coherence in depth estimation, outperforming prior work like MapAnything.
DrivingDepth addresses the geometry-scale conflict in driving depth estimation by using sparse LiDAR as prompts to locally calibrate a frozen foundation model via per-pixel scale correction, achieving state-of-the-art metric accuracy (AbsRel 11.19) and geometric consistency (EdgeCR 5.741) on nuScenes.
Dense depth estimation for autonomous driving faces a geometry-scale conflict: depth foundation models deliver pixel-aligned dense visual geometry without reliable metric scale, while projected LiDAR provides metric anchors that are sparse, noisy, and misaligned with image structures. Existing sparse-prompted methods incorporate LiDAR by regenerating depth from scratch, overriding the foundation model's coherent geometry and producing structural artifacts on visually continuous surfaces. Our key insight is that foundation models already capture geometrically coherent relative depth; no additional surface structure learning is required-only a per-pixel scale factor mapping relative geometry to metric coordinates. Based on this, we propose DrivingDepth, which treats sparse LiDAR as geometric prompts that locally calibrate a frozen foundation prior through residual pixel-wise scale correction, preserving dense visual geometry by construction. On nuScenes with 4-frame surround-view input, DrivingDepth achieves an AbsRel of 11.19 and an EdgeCR of 5.741, outperforming MapAnything (11.99/1.914) by simultaneously delivering SOTA metric accuracy and geometric consistency.