Sim-to-Real Fruit Detection Using Synthetic Data: Quantitative Evaluation and Embedded Deployment with Isaac Sim

View PDF HTML (experimental)

Abstract:This study investigates the effectiveness of synthetic data for sim-to-real transfer in object detection under constrained data conditions and embedded deployment requirements. Synthetic datasets were generated in NVIDIA Isaac Sim and combined with limited real-world fruit images to train YOLO-based detection models under real-only, synthetic-only, and hybrid regimes. Performance was evaluated on two test datasets: an in-domain dataset with conditions matching the training data and a domain shift dataset containing real fruit and different background conditions. Results show that models trained exclusively on real data achieve the highest accuracy, while synthetic-only models exhibit reduced performance due to a domain gap. Hybrid training strategies significantly improve performance compared to synthetic-only approaches and achieve results close to real-only training while reducing the need for manual annotation. Under domain shift conditions, all models show performance degradation, with hybrid models providing improved robustness. The trained models were successfully deployed on a Jetson Orin NX using TensorRT optimization, achieving real-time inference performance. The findings highlight that synthetic data is most effective when used in combination with real data and that deployment constraints must be considered alongside detection accuracy.

Comments: 18 pages, 6 figures

Subjects:

Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)

Cite as: arXiv:2603.28670 [cs.CV]

(or arXiv:2603.28670v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2603.28670

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Martina Hutter-Mironovova [view email] [v1] Mon, 30 Mar 2026 16:52:29 UTC (2,037 KB)