Mind the Domain Gap: Measuring the Domain Gap Between Real-World and Synthetic Point Clouds for Automated Driving Development

2025-05-23

AI Summary

Quantifying the domain gap between real and synthetic point clouds in autonomous driving remains challenging, leading to degraded perception performance and safety risks. Method: This paper proposes the first geometry-semantic joint domain discrepancy assessment method tailored to co-geographic scenes. It introduces DoGSS-PCL—a novel metric that disentangles three sources of domain shift: network capability, class definition, and object representation—integrating point cloud distribution comparison, semantic consistency verification, multi-scale geometric modeling, and differentiable rendering validation. Results: Experiments demonstrate that perception models maintain performance using only 50% real and 50% synthetic point clouds. This work overcomes limitations of conventional cross-scene evaluation, establishing the first credible simulation assessment standard for digital twin testing and significantly improving robustness validation efficiency.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Diffusion Models for Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Security and Privacy: Large-scale security measurements

Abstract

Owing to the typical long-tail data distribution issues, simulating domain-gap-free synthetic data is crucial in robotics, photogrammetry, and computer vision research. The fundamental challenge pertains to credibly measuring the difference between real and simulated data. Such a measure is vital for safety-critical applications, such as automated driving, where out-of-domain samples may impact a car's perception and cause fatal accidents. Previous work has commonly focused on simulating data on one scene and analyzing performance on a different, real-world scene, hampering the disjoint analysis of domain gap coming from networks' deficiencies, class definitions, and object representation. In this paper, we propose a novel approach to measuring the domain gap between the real world sensor observations and simulated data representing the same location, enabling comprehensive domain gap analysis. To measure such a domain gap, we introduce a novel metric DoGSS-PCL and evaluation assessing the geometric and semantic quality of the simulated point cloud. Our experiments corroborate that the introduced approach can be used to measure the domain gap. The tests also reveal that synthetic semantic point clouds may be used for training deep neural networks, maintaining the performance at the 50/50 real-to-synthetic ratio. We strongly believe that this work will facilitate research on credible data simulation and allow for at-scale deployment in automated driving testing and digital twinning.

Problem

• Measure domain gap between real and synthetic point clouds
• Assess geometric and semantic quality of simulated data
• Enable credible data simulation for automated driving

Innovation

• Measure domain gap between real and synthetic data
• Introduce DoGSS-PCL metric for evaluation
• Enable training deep networks with synthetic data