Perfecting Depth: Uncertainty-Aware Enhancement of Metric Depth

2025-06-05

AI Summary

To address artifacts and accuracy degradation caused by unreliable regions in sensor-derived depth maps, this paper proposes a two-stage collaborative enhancement framework. In the first stage, a diffusion model leverages stochasticity to characterize the training-inference domain gap, enabling unsupervised reliability awareness and automatic detection of unreliable regions. In the second stage, geometric consistency constraints are integrated with deterministic fine-tuning to enhance depth map density and metric accuracy while preserving structural integrity. This work introduces the novel paradigm of “stochastic uncertainty modeling + deterministic structural constraint” collaboration and is the first to achieve domain-gap-driven reliability awareness without supervision from reliable ground truth. Experiments demonstrate that the resulting depth maps are artifact-free, highly dense, and exhibit significantly improved reliability over state-of-the-art methods, establishing a new benchmark for sensor depth enhancement.

Technology Category

• Reasoning under Uncertainty: Stochastic Optimization
• Computer Vision: Diffusion Models for Vision
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

We propose a novel two-stage framework for sensor depth enhancement, called Perfecting Depth. This framework leverages the stochastic nature of diffusion models to automatically detect unreliable depth regions while preserving geometric cues. In the first stage (stochastic estimation), the method identifies unreliable measurements and infers geometric structure by leveraging a training-inference domain gap. In the second stage (deterministic refinement), it enforces structural consistency and pixel-level accuracy using the uncertainty map derived from the first stage. By combining stochastic uncertainty modeling with deterministic refinement, our method yields dense, artifact-free depth maps with improved reliability. Experimental results demonstrate its effectiveness across diverse real-world scenarios. Furthermore, theoretical analysis, various experiments, and qualitative visualizations validate its robustness and scalability. Our framework sets a new baseline for sensor depth enhancement, with potential applications in autonomous driving, robotics, and immersive technologies.

Problem

• Enhancing sensor depth maps for reliability and accuracy
• Detecting unreliable depth regions using diffusion models
• Improving depth maps for autonomous driving and robotics

Innovation

• Two-stage framework for depth enhancement
• Leverages diffusion models for uncertainty detection
• Combines stochastic and deterministic refinement