You Only Flow Once: Calibrated and Real-Time Radar Pose Estimation with Multi-Hypothesis Normalizing Flows
2026-08-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors point out that millimeter-wave radar data is often messy and unclear, making it hard to figure out exactly how a person is posing since many different poses might fit the data. Current methods try to give just one guess, which can be wrong. They introduce a new method called MH-NFPG that uses a special flow model combined with a transformer to quickly generate multiple possible poses at once, with better accuracy and reliable uncertainty estimates. Their method works faster and is more dependable than previous diffusion-based methods across several tests. This shows their approach is practical for real-time human pose estimation using radar data.
millimeter-wave radarpose estimationnormalizing flowsconditional normalizing flowspatiotemporal transformerdiffusion modelsuncertainty calibrationmulti-hypothesis modelingposterior distributionreal-time inference
Authors
Jonas Leo Mueller, Sebastian Hoefler, Dario Zanca, Naga Venkata Sai Jitin Jami, Thomas Altstidl, Bjoern M. Eskofier
Abstract
Sparse and noisy millimeter-wave radar point cloud observations often correspond to multiple plausible human poses, making deterministic pose estimation fundamentally ill-posed. Yet existing radar methods remain deterministic, collapsing this ambiguity into a single estimate. Diffusion-based alternatives can model multi-hypothesis distributions but require costly sequential denoising for each distribution sample and lack calibrated uncertainty. We propose Multi-Hypothesis Normalizing Flow Pose Generator (MH-NFPG), which models pose distributions from radar point clouds using a conditional normalizing flow. Specifically, we combine a spatiotemporal transformer backbone with a normalizing flow that transforms a Laplace base distribution into an expressive posterior, generated in parallel through a single forward pass. Leveraging this efficiency, we outperform diffusion-based alternatives in calibration across three radar benchmarks (MM-Fi, mmRadPose, mRI), improve pose accuracy on two, and match it on the third, while achieving over 20x faster inference for applications and reducing calibration error by up to 85%. We find that calibration degrades substantially for diffusion models, whereas our flow-based approach maintains reliable coverage, also in cross-environment settings. These results demonstrate normalizing flows as a practical alternative to diffusion models for real-time, uncertainty-aware radar pose estimation. Our code will be made publicly available.