TransNormal: Dense Visual Semantics for Diffusion-based Transparent Object Normal Estimation

2026-01-31

AI Summary

Monocular normal estimation for transparent objects remains challenging due to distortions caused by refraction and reflection, which hinder accurate capture by conventional sensors and limit the deployment of embodied intelligence in scientific environments. To address this, we propose TransNormal, a novel framework that, for the first time, integrates pre-trained diffusion priors with dense visual semantics from DINOv3 to guide single-step normal regression via cross-attention mechanisms. The approach further incorporates multi-task learning and wavelet-based regularization to preserve fine-grained geometric details. We also introduce TransNormal-Synthetic, the first high-fidelity synthetic dataset of transparent laboratory glassware. Evaluated on ClearGrasp and ClearPose, our method reduces mean angular error by 24.4% and 15.2%, respectively, and improves the 11.25° accuracy metric by 22.8%, substantially outperforming existing approaches.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Security and Privacy: Data transparency and provenance
• Responsible Web: Algorithmic accountability and transparency on the web
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization

Abstract

Monocular normal estimation for transparent objects is critical for laboratory automation, yet it remains challenging due to complex light refraction and reflection. These optical properties often lead to catastrophic failures in conventional depth and normal sensors, hindering the deployment of embodied AI in scientific environments. We propose TransNormal, a novel framework that adapts pre-trained diffusion priors for single-step normal regression. To handle the lack of texture in transparent surfaces, TransNormal integrates dense visual semantics from DINOv3 via a cross-attention mechanism, providing strong geometric cues. Furthermore, we employ a multi-task learning objective and wavelet-based regularization to ensure the preservation of fine-grained structural details. To support this task, we introduce TransNormal-Synthetic, a physics-based dataset with high-fidelity normal maps for transparent labware. Extensive experiments demonstrate that TransNormal significantly outperforms state-of-the-art methods: on the ClearGrasp benchmark, it reduces mean error by 24.4% and improves 11.25{\deg} accuracy by 22.8%; on ClearPose, it achieves a 15.2% reduction in mean error. The code and dataset will be made publicly available at https://longxiang-ai.github.io/TransNormal.

Problem

• transparent object
• normal estimation
• monocular vision
• light refraction
• embodied AI

Innovation

• diffusion prior
• dense visual semantics
• cross-attention
• multi-task learning
• wavelet regularization