Identify the primary causes of the synthetic-to-real domain gap

Identify the principal factors that cause the performance gap between synthetic egocentric hand–object interaction data and real-world data for hand–object interaction detection.

Background

Understanding the sources of the domain gap is essential for designing targeted improvements in simulation (e.g., grasp realism, environment diversity) and modeling (e.g., adaptation techniques).

By diagnosing the causes, researchers can prioritize simulator enhancements and training strategies that most effectively reduce the discrepancy between synthetic and real data.

References

As a result, several key open questions still need to be addressed: 1) How large is the gap between synthetic and real data? 2) What are its main causes?

Leveraging Synthetic Data for Enhancing Egocentric Hand-Object Interaction Detection  (2603.29733 - Leonardi et al., 31 Mar 2026) in Section 1 (Introduction)