arXiv subjects

Jonas Emrich

Publications and source records attributed to Jonas Emrich.

2 recordsLinked to original sources

arXiv2026

A magnetic environment with reproducible spatio-temporal magnetic conditions at picotesla level

Walk-in, picotesla-scale environments are essential for measurements of biomagnetism and fundamental physics. However, conventional rooms require typical active dynamic compensation, which adds complexity and magnetic noise. Here, we present a solution that achieves an absolute residual field well below 100$\,$pT within its central measurement volume. Following magnetic equilibration, this environment achieves picotesla-scale reproducibility. Consequently, optically pumped magnetometers operate at their design noise and drift performance and remain operational during sensor motion without active feedback. A key technique is robotic mapping, which resolves the ultra-low residual field patterns and demagnetization stability. We demonstrate the platform's versatility through high-fidelity adult and fetal magnetocardiography, standing magnetoencephalography, and ultra-low field magnetic resonance with polarized noble gases in the limit of strongly coupled spins in a negligible holding field. The achieved passive reproducibility turns the ultra-low magnetic background into a predictable, correctable property, establishing a novel foundation for next-generation quantum sensing and precision physics.

physics.ins-det↗

arXiv2025

MeshFleet: Filtered and Annotated 3D Vehicle Dataset for Domain Specific Generative Modeling

Generative models have recently made remarkable progress in the field of 3D objects. However, their practical application in fields like engineering remains limited since they fail to deliver the accuracy, quality, and controllability needed for domain-specific tasks. Fine-tuning large generative models is a promising perspective for making these models available in these fields. Creating high-quality, domain-specific 3D datasets is crucial for fine-tuning large generative models, yet the data filtering and annotation process remains a significant bottleneck. We present MeshFleet, a filtered and annotated 3D vehicle dataset extracted from Objaverse-XL, the most extensive publicly available collection of 3D objects. Our approach proposes a pipeline for automated data filtering based on a quality classifier. This classifier is trained on a manually labeled subset of Objaverse, incorporating DINOv2 and SigLIP embeddings, refined through caption-based analysis and uncertainty estimation. We demonstrate the efficacy of our filtering method through a comparative analysis against caption and image aesthetic score-based techniques and fine-tuning experiments with SV3D, highlighting the importance of targeted data selection for domain-specific 3D generative modeling.

cs.CV↗