A Neural Quality Metric for BRDF Models

2025-08-04

AI Summary

Traditional BRDF quality assessment relies on numerical error metrics, which poorly correlate with human perception of material appearance. To address this, we propose the first perceptually grounded neural metric operating directly in BRDF space—capable of predicting visually detectable differences between reference and approximated BRDFs without rendering. Our method pioneers the transfer of image-space perceptual knowledge to the BRDF domain, implementing a lightweight multi-layer perceptron (MLP) that takes paired reference and approximate BRDFs as input and outputs a Just-Noticeable Difference (JOD) score. The model is trained on a hybrid dataset comprising both measured and synthetically generated BRDFs. Extensive experiments demonstrate that our metric achieves significantly higher correlation with human subjective judgments than existing BRDF-space metrics, establishing a new, perception-aligned evaluation benchmark for material modeling.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Evaluation and Analysis
• Humans and AI: Brain-Sensing and Analysis

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Accurately evaluating the quality of bidirectional reflectance distribution function (BRDF) models is essential for photo-realistic rendering. Traditional BRDF-space metrics often employ numerical error measures that fail to capture perceptual differences evident in rendered images. In this paper, we introduce the first perceptually informed neural quality metric for BRDF evaluation that operates directly in BRDF space, eliminating the need for rendering during quality assessment. Our metric is implemented as a compact multi-layer perceptron (MLP), trained on a dataset of measured BRDFs supplemented with synthetically generated data and labelled using a perceptually validated image-space metric. The network takes as input paired samples of reference and approximated BRDFs and predicts their perceptual quality in terms of just-objectionable-difference (JOD) scores. We show that our neural metric achieves significantly higher correlation with human judgments than existing BRDF-space metrics. While its performance as a loss function for BRDF fitting remains limited, the proposed metric offers a perceptually grounded alternative for evaluating BRDF models.

Problem

• Develop a neural metric for BRDF quality evaluation
• Replace traditional numerical error measures with perceptual accuracy
• Predict perceptual quality using JOD scores without rendering

Innovation

• Perceptually informed neural BRDF quality metric
• Compact MLP trained on measured and synthetic BRDFs
• Predicts perceptual quality using JOD scores