The Power of Light: Improving Synthetic-to-Real Domain Adaptation through Physically-Based Indirect Illumination

2026-06-21

AI Summary

This work addresses the performance degradation in object detection caused by domain gaps between synthetic and real-world data in industrial automation. To bridge this gap, the authors propose SmartSDG, a physics-based automated synthetic data generation pipeline, and introduce ILLUM_INTRUCK, a new industrial benchmark dataset. Leveraging NVIDIA Isaac Sim with physically based rendering (PBR) and integrated with the YOLOv12 detection framework, the study systematically quantifies the impact of indirect illumination and background complexity on domain adaptation, revealing their critical roles for the first time. Experimental results demonstrate that, compared to conventional direct-illumination synthetic data, SmartSDG better preserves surface textures, reduces false positives, accelerates model convergence, and significantly narrows the performance gap between synthetic and real domains, while offering actionable guidelines for virtual scene design.

Abstract

While synthetic data generation resolves the manual labeling bottleneck in computer vision, minimizing the syn-to-real domain gap requires optimizing rendering variables. This paper presents a systematic study analyzing the impact of lighting configurations and background complexity on object detection performance. We introduce SmartSDG, an automated, reproducible pipeline built on NVIDIA Isaac Sim using Physically-Based Shading (PBS), alongside ILLUM\_INTRUCK, a new multi-object industrial benchmark dataset. Through 18 controlled experiments utilizing a state-of-the-art YOLOv12 framework, we demonstrate that complex, indirect lighting configurations paired with domain-relevant background variability significantly increase visual cue richness. Our quantitative findings show that avoiding direct specular peaks preserves crucial surface textures, mitigates the domain gap, reduces false positives, and accelerates model convergence compared to using conventional direct-light synthetic data. Ultimately, we provide actionable virtual scene design guidelines to maximize object detection robustness in industrial automation.

Problem

• domain adaptation
• synthetic-to-real gap
• indirect illumination
• object detection
• physically-based rendering

Innovation

• Physically-Based Shading
• Indirect Illumination
• Synthetic-to-Real Domain Adaptation
• SmartSDG
• Industrial Object Detection