GVDepth: Zero-Shot Monocular Depth Estimation for Ground Vehicles based on Probabilistic Cue Fusion

TL;DR

GVDepth introduces a probabilistic cue fusion method for zero-shot monocular depth estimation in ground vehicles, leveraging fixed camera-ground relationships to improve generalization across diverse datasets and camera configurations.

Contribution

The paper presents a novel canonical representation and a probabilistic fusion architecture that enhance zero-shot depth estimation for ground vehicles across multiple datasets.

Findings

Achieves accurate depth estimation across five autonomous driving datasets.

Performs comparably to existing zero-shot methods despite single-dataset training.

Effectively generalizes across different camera setups and resolutions.

Abstract

Generalizing metric monocular depth estimation presents a significant challenge due to its ill-posed nature, while the entanglement between camera parameters and depth amplifies issues further, hindering multi-dataset training and zero-shot accuracy. This challenge is particularly evident in autonomous vehicles and mobile robotics, where data is collected with fixed camera setups, limiting the geometric diversity. Yet, this context also presents an opportunity: the fixed relationship between the camera and the ground plane imposes additional perspective geometry constraints, enabling depth regression via vertical image positions of objects. However, this cue is highly susceptible to overfitting, thus we propose a novel canonical representation that maintains consistency across varied camera setups, effectively disentangling depth from specific parameters and enhancing generalization across datasets. We also propose a novel architecture that adaptively and probabilistically fuses depths estimated via object size and vertical image position cues. A comprehensive evaluation demonstrates the effectiveness of the proposed approach on five autonomous driving datasets, achieving accurate metric depth estimation for varying resolutions, aspect ratios and camera setups. Notably, we achieve comparable accuracy to existing zero-shot methods, despite training on a single dataset with a single-camera setup. Project website: https://unizgfer-lamor.github.io/gvdepth/