Depth Anything at Any Condition

TL;DR

Depth Anything at Any Condition (DepthAnything-AC) is a versatile monocular depth estimation model that excels in diverse and challenging environments by using unsupervised finetuning and spatial constraints, demonstrating strong zero-shot performance.

Contribution

The paper introduces an unsupervised finetuning paradigm and spatial distance constraint to improve depth estimation in complex conditions without extensive labeled data.

Findings

Achieves zero-shot depth estimation across various benchmarks.

Performs well in adverse weather and corrupted environments.

Provides clear semantic boundaries and detailed depth maps.

Abstract

We present Depth Anything at Any Condition (DepthAnything-AC), a foundation monocular depth estimation (MDE) model capable of handling diverse environmental conditions. Previous foundation MDE models achieve impressive performance across general scenes but not perform well in complex open-world environments that involve challenging conditions, such as illumination variations, adverse weather, and sensor-induced distortions. To overcome the challenges of data scarcity and the inability of generating high-quality pseudo-labels from corrupted images, we propose an unsupervised consistency regularization finetuning paradigm that requires only a relatively small amount of unlabeled data. Furthermore, we propose the Spatial Distance Constraint to explicitly enforce the model to learn patch-level relative relationships, resulting in clearer semantic boundaries and more accurate details. Experimental results demonstrate the zero-shot capabilities of DepthAnything-AC across diverse benchmarks, including real-world adverse weather benchmarks, synthetic corruption benchmarks, and general benchmarks. Project Page: https://ghost233lism.github.io/depthanything-AC-page Code: https://github.com/HVision-NKU/DepthAnythingAC