LiDAR Prompted Spatio-Temporal Multi-View Stereo for Autonomous Driving

TL;DR

DriveMVS is a multi-view stereo framework that integrates LiDAR prompts and spatio-temporal cues to achieve high-accuracy, consistent depth estimation for autonomous driving, excelling in cross-domain scenarios.

Contribution

It introduces a novel method that uses LiDAR as a geometric prompt and a spatio-temporal decoder to improve depth accuracy and consistency in autonomous driving perception.

Findings

Achieves state-of-the-art results on multiple benchmarks.

Demonstrates superior temporal stability and cross-domain transfer.

Effectively integrates LiDAR prompts with deep fusion techniques.

Abstract

Accurate metric depth is critical for autonomous driving perception and simulation, yet current approaches struggle to achieve high metric accuracy, multi-view and temporal consistency, and cross-domain generalization. To address these challenges, we present DriveMVS, a novel multi-view stereo framework that reconciles these competing objectives through two key insights: (1) Sparse but metrically accurate LiDAR observations can serve as geometric prompts to anchor depth estimation in absolute scale, and (2) deep fusion of diverse cues is essential for resolving ambiguities and enhancing robustness, while a spatio-temporal decoder ensures consistency across frames. Built upon these principles, DriveMVS embeds the LiDAR prompt in two ways: as a hard geometric prior that anchors the cost volume, and as soft feature-wise guidance fused by a triple-cue combiner. Regarding temporal consistency, DriveMVS employs a spatio-temporal decoder that jointly leverages geometric cues from the MVS cost volume and temporal context from neighboring frames. Experiments show that DriveMVS achieves state-of-the-art performance on multiple benchmarks, excelling in metric accuracy, temporal stability, and zero-shot cross-domain transfer, demonstrating its practical value for scalable, reliable autonomous driving systems.