DarkVGGT: Seeing Through Darkness Using Thermal Geometry without Daylight Tax

2026-06-09

AI Summary

This work addresses the significant performance degradation of visible-light 3D reconstruction in dark and low-visibility conditions due to deteriorated RGB signals. To overcome this limitation, the authors propose DarkVGGT, a feed-forward geometric framework that fuses RGB and thermal imaging through physics-aware thermal modeling to enable robust 3D geometry estimation. The core innovations include a thermal-radiation-driven mechanism for extracting geometrically consistent cues, a cross-modal routing strategy that shares geometric structures between modalities, and a physics-inspired thermal factorization method that effectively suppresses reflective interference while injecting reliability-aware structural guidance. Experiments demonstrate that DarkVGGT substantially outperforms existing feed-forward geometric approaches on low-visibility RGB-T benchmarks, achieving state-of-the-art accuracy in both depth estimation and camera pose recovery, while maintaining strong performance under normal lighting conditions.

Abstract

Recent feed-forward 3D reconstruction methods have demonstrated strong performance and flexibility in efficient end-to-end scene geometry estimation from image streams. However, their reliance on visible-light appearance makes them vulnerable in dark and low-visibility environments, where RGB cues are severely degraded and geometric evidence becomes ambiguous. To address this challenge, we propose DarkVGGT, an RGB-T feed-forward geometry framework that uses physics-aware thermal modeling for robust 3D estimation in low-light scenes. DarkVGGT introduces two complementary modules. First, physics-inspired thermal factorization extracts emissive-dominant, geometry-consistent thermal cues while isolating sparse reflective residuals that may introduce geometric ambiguity. Second, geometry-shared thermal routing isolates modality-invariant geometric structures from thermal-specific patterns, selectively injecting reliability-aware structural guidance into the RGB stream. Together, these components enable accurate thermal-informed geometry estimation under degraded RGB conditions while largely preserving performance in well-lit environments. Experiments on low-visibility RGB-T benchmarks demonstrate consistent improvements in both depth and camera pose estimation over existing feed-forward geometry baselines.

Problem

• 3D reconstruction
• low-light environments
• thermal imaging
• geometric ambiguity
• RGB-T

Innovation

• thermal geometry
• RGB-T fusion
• low-light 3D reconstruction
• physics-aware modeling
• feed-forward geometry