StableIntrinsic: Detail-preserving One-step Diffusion Model for Multi-view Material Estimation

2025-08-27

AI Summary

Existing multi-step denoising diffusion models for multi-view material estimation suffer from high inference latency and large output variance, hindering deterministic material recovery. This paper proposes StableIntrinsic—the first end-to-end, single-step diffusion model specifically designed for intrinsic material estimation. Its key contributions are: (1) a deterministic single-step diffusion architecture eliminating stochastic iterative sampling; (2) a material-aware pixel-wise multi-task loss function jointly optimizing albedo, metallic, and roughness; and (3) a Detail Injection Network (DIN) that mitigates texture degradation induced by VAE encoding. Trained on multi-view RGB images, StableIntrinsic achieves state-of-the-art performance: +9.9 dB PSNR on albedo, and 44.4% and 60.0% reductions in MSE for metallic and roughness maps, respectively—significantly outperforming prior methods.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Intelligent Robots: State Estimation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Recovering material information from images has been extensively studied in computer graphics and vision. Recent works in material estimation leverage diffusion model showing promising results. However, these diffusion-based methods adopt a multi-step denoising strategy, which is time-consuming for each estimation. Such stochastic inference also conflicts with the deterministic material estimation task, leading to a high variance estimated results. In this paper, we introduce StableIntrinsic, a one-step diffusion model for multi-view material estimation that can produce high-quality material parameters with low variance. To address the overly-smoothing problem in one-step diffusion, StableIntrinsic applies losses in pixel space, with each loss designed based on the properties of the material. Additionally, StableIntrinsic introduces a Detail Injection Network (DIN) to eliminate the detail loss caused by VAE encoding, while further enhancing the sharpness of material prediction results. The experimental results indicate that our method surpasses the current state-of-the-art techniques by achieving a $9.9%$ improvement in the Peak Signal-to-Noise Ratio (PSNR) of albedo, and by reducing the Mean Square Error (MSE) for metallic and roughness by $44.4%$ and $60.0%$, respectively.

Problem

• Addressing multi-step diffusion inefficiency in material estimation
• Reducing variance in deterministic material parameter prediction
• Solving detail loss from VAE encoding in diffusion models

Innovation

• One-step diffusion model for material estimation
• Detail Injection Network enhances prediction sharpness
• Losses in pixel space based on material properties