ARDepth: Auto-regressive Monocular Depth Estimation with Progressive Visual Conditioning

2026-07-14 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

ARDepth introduces a novel approach for monocular depth estimation (MDE), formulating it as structured auto-regressive generation. This method contrasts with diffusion models by progressively constructing depth representations as spatial resolution increases, rather than relying on global iterative denoising. To achieve this, ARDepth incorporates Scale-Progressive Conditioning (SPC) to inject multi-scale visual features at each generation stage and Semantic-Aware Guidance (SAG) to provide scene-level semantic priors. These designs enable the model to capture fine-grained local details while maintaining coherent global geometry. Empirical results demonstrate ARDepth's strong performance and its ability to produce structurally consistent depth predictions across scales.

Key takeaway

For Machine Learning Engineers developing monocular depth estimation solutions, ARDepth presents a compelling alternative to diffusion models by leveraging auto-regressive generation. You should explore auto-regressive approaches for MDE, especially when fine-grained local details and global structural consistency are critical. Consider integrating multi-scale conditioning and semantic guidance into your model architectures to enhance performance and coherence.

Key insights

Auto-regressive generation offers a promising paradigm for monocular depth estimation, progressively building depth representations for structural consistency.

Principles

• Depth estimation benefits from progressive construction across spatial scales.
• Multi-scale visual features enhance generative depth models.
• Semantic priors improve global structural consistency in MDE.

Method

ARDepth formulates MDE as structured auto-regressive generation, progressively constructing depth representations with increasing spatial resolution using Scale-Progressive Conditioning (SPC) and Semantic-Aware Guidance (SAG).

In practice

• Consider auto-regressive generation for geometric modeling tasks.
• Integrate multi-scale conditioning in generative models.
• Apply semantic guidance for structural coherence.

Topics

• Monocular Depth Estimation
• Auto-regressive Models
• Computer Vision
• Generative Models
• Scale-Progressive Conditioning
• Semantic-Aware Guidance

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• ARDY: Autoregressive Diffusion with Hybrid Representation for Interactive Human Motion Generation — ARDY enables real-time, high-fidelity 3D human motion generation with online text and kinematic control.
• Modality Forcing for Scalable Spatial Generation — Modality Forcing enables scalable, joint image-depth generation from sparse data by employing T2I priors with a simple post-training recipe.
• JustDepth: Real-Time Radar-Camera Depth Estimation with Single-Scan LiDAR Supervision — Single-stage radar-camera depth estimation achieves real-time performance and artifact reduction with minimal supervision.
• Rethinking Text-to-Image as Semantic-Aware Data Augmentation for Indoor Scene Recognition — Using Stable Diffusion for semantic-aware data augmentation effectively addresses indoor scene recognition data scarcity and enables robust synthetic image detection.
• Diffusion LLMs from the Ground Up: Training, Inference, and Practical Engineering — dLLMs offer competitive performance and significantly faster inference than AR models, especially when initialized from pre-trained AR checkpoints.
• The 3D Mirage: Probing and Taming 3D Hallucinations — Monocular depth models hallucinate 3D structures from 2D illusions, requiring targeted benchmarks and mitigation.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.