Distill Any Depth: Distillation Creates a Stronger Monocular Depth Estimator

TL;DR

This paper improves zero-shot monocular depth estimation by analyzing depth normalization strategies, proposing a cross-context distillation method that combines global and local cues, and integrating diffusion-based priors for more robust supervision.

Contribution

It introduces a systematic analysis of depth normalization in pseudo-label distillation and proposes Cross-Context Distillation with diffusion-based priors to enhance depth estimation.

Findings

Outperforms state-of-the-art methods on benchmark datasets.

Omitting normalization can reduce noise impact in recent distillation paradigms.

Combining global and local cues improves pseudo-label quality.

Abstract

Recent advances in zero-shot monocular depth estimation(MDE) have significantly improved generalization by unifying depth distributions through normalized depth representations and by leveraging large-scale unlabeled data via pseudo-label distillation. However, existing methods that rely on global depth normalization treat all depth values equally, which can amplify noise in pseudo-labels and reduce distillation effectiveness. In this paper, we present a systematic analysis of depth normalization strategies in the context of pseudo-label distillation. Our study shows that, under recent distillation paradigms (e.g., shared-context distillation), normalization is not always necessary, as omitting it can help mitigate the impact of noisy supervision. Furthermore, rather than focusing solely on how depth information is represented, we propose Cross-Context Distillation, which integrates both global and local depth cues to enhance pseudo-label quality. We also introduce an assistant-guided distillation strategy that incorporates complementary depth priors from a diffusion-based teacher model, enhancing supervision diversity and robustness. Extensive experiments on benchmark datasets demonstrate that our approach significantly outperforms state-of-the-art methods, both quantitatively and qualitatively.