Synthetisierte Objekte in Radardaten zum Trainieren oder Validieren eines Maschinenlernmodells

Details

Veröffentlichungs-Nr.

EP4800430

Anmeldetag

27. Februar 2025

Veröffentlichung

2. September 2026

Rechtsraum

EP

Abstract

Techniques for populating a dataset for training or validating a machine-learning model are disclosed. One or more measurement samples (811, 812) are obtained, each of the one or more measurement samples comprising respective radar data (120, 140, 420) observing a respective scene (170, 171) comprising one or more respective objects (181, 182). A reflection signal component of a selected object from the respective radar data and synthesized radar data (821) is then generated based on a combination of the this reflection signal component acting as a template and at least a part of the radar data of the one or more measurement samples. Thereby, the radar data (821) emulates an observation of a synthesized scene (172) comprising at least the synthesized object (183). A synthesized sample comprising the synthesized radar data is then added to the dataset.

Anmelder

Firma

Infineon Technologies AG

Land

🇩🇪 Deutschland