SynDRA-Gen: A Configurable Pipeline for Controlled Synthetic Multi-Modal Data Generation in a Railway Shunting Yard

Le résumé fourni par la source

The development of perception systems for railway applications relies on large volumes of annotated multimodal data.However, acquiring such data in real-world railway environments is inherently challenging due to safety constraints, high operational costs, and limited access to infrastructure.This paper introduces SynDRA-Gen, a standalone application designed for controlled synthetic data generation in railway yard scenarios.The system builds upon a previously developed simulation framework used for railway perception datasets and extends it into a fully configurable and user-friendly pipeline.SynDRA-Gen enables users to customize scene layouts, environmental conditions, and sensor configurations through a structured interface.The application supports a synchronized sensor suite that generates both camera and LiDAR data, along with automatically produced annotations such as depth maps, semantic segmentation, and 2D/3D bounding boxes.To facilitate reproducibility and accessibility, the system is distributed as a precompiled executable, allowing consistent dataset generation without requiring direct access to the underlying simulation framework.The tool is publicly available at: https://github.com/DAmicoGlc/SynDRA-Gen.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.