3D Blood Pulsation Maps

2025-12-11

AI Summary

Existing research is hindered by the lack of multi-view physiological video datasets supporting 3D pulsatile blood flow modeling, limiting the robustness and interpretability of remote photoplethysmography imaging (PPGI). To address this, we introduce Pulse3DFace—the first multi-view facial video dataset designed for 3D pulsatile modeling—comprising RGB videos (30 Hz) from 23 viewpoints across 15 subjects. We reconstruct high-fidelity 3D facial geometry via Structure-from-Motion (SfM) and, for the first time, provide FLAME-texture-aligned 3D pulsation maps encoding physiologically plausible spatiotemporal features, including signal-to-noise ratio, local amplitude, and phase. The dataset is rigorously synchronized with ground-truth ECG signals, demonstrating exceptional consistency under varying illumination conditions. It accurately captures coordinated pulsatile dynamics across the face and neck, thereby substantially improving PPGI algorithm generalization and enabling mechanistic analysis of physiological signal propagation.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Data transparency and provenance

Abstract

We present Pulse3DFace, the first dataset of its kind for estimating 3D blood pulsation maps. These maps can be used to develop models of dynamic facial blood pulsation, enabling the creation of synthetic video data to improve and validate remote pulse estimation methods via photoplethysmography imaging. Additionally, the dataset facilitates research into novel multi-view-based approaches for mitigating illumination effects in blood pulsation analysis. Pulse3DFace consists of raw videos from 15 subjects recorded at 30 Hz with an RGB camera from 23 viewpoints, blood pulse reference measurements, and facial 3D scans generated using monocular structure-from-motion techniques. It also includes processed 3D pulsation maps compatible with the texture space of the 3D head model FLAME. These maps provide signal-to-noise ratio, local pulse amplitude, phase information, and supplementary data. We offer a comprehensive evaluation of the dataset's illumination conditions, map consistency, and its ability to capture physiologically meaningful features in the facial and neck skin regions.

Problem

• Estimates 3D blood pulsation maps from facial videos
• Enables synthetic video generation for remote pulse validation
• Facilitates multi-view approaches to reduce illumination effects

Innovation

• Dataset for 3D blood pulsation maps
• Multi-view video to reduce illumination effects
• Processed maps compatible with 3D head model