Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.
Lightweight, physics-motivated rPPG. The authors model the pulse as a second-order dynamical system (from Navier–Stokes hemodynamics) and argue its discrete solution is a causal convolution, which justifies a temporal convolutional network. The network adds two cheap spatial pieces: a parameter-free Zero-FLOPs Axial Swapper that permutes a few channels to mix distant face regions, and a learned per-frame soft spatial mask.
ESTBlock spatio-temporal encoder with Zero-FLOPs Axial Swapper (parameter-free channel/axis permutation mixing distant face regions) → Adaptive Spatial Filter (per-frame soft spatial mask) → Gated TCN (causal dilated TCN with gating, 3 layers default); loss weight λ = 0.1
Remote photoplethysmography (rPPG) measurement enables non-contact physiological monitoring but suffers from accuracy degradation under head motion and illumination changes. Existing deep learning methods are mostly heuristic and lack theoretical grounding, limiting robustness and interpretability. In this work, we propose a physics-informed rPPG paradigm derived from the Navier-Stokes equations of hemodynamics, showing that the pulse signal follows a second-order dynamical system whose discrete solution naturally leads to a causal convolution, justifying the use of a Temporal Convolutional Network (TCN). Based on this principle, we design the PHASE-Net, a lightweight model with three key components: 1) Zero-FLOPs Axial Swapper module to swap or transpose a few spatial channels to mix distant facial regions, boosting cross-region feature interaction without changing temporal order; 2) Adaptive Spatial Filter to learn a soft spatial mask per frame to highlight signal-rich areas and suppress noise for cleaner feature maps; and 3) Gated TCN, a causal dilated TCN with gating that models long-range temporal dynamics for accurate pulse recovery. Extensive experiments demonstrate that PHASE-Net achieves state-of-the-art performance and strong efficiency, offering a theoretically grounded and deployment-ready rPPG solution. The source code is available at https://github.com/Alex036225/PhaseNet.
Verbatim fromarXiv· arXiv metadata (abstracts are released under CC0)
Datasets linked in the hub; pretraining vs evaluation roles follow the paper.
| Task | Dataset | Metric | Value | Rank |
|---|---|---|---|---|
| Rppg hr | BUAA-MIHR intra | HR MAE ↓ | 5.89 | #3 / 4 |
| Rppg hr | PURE intra | HR MAE ↓ | 0.14 | #1 / 5 |
| Rppg hr | UBFC-rPPG intra | HR MAE ↓ | 0.15 | #3 / 5 |
| Rppg hr | mmpd intra | HR MAE ↓ | 4.78 | #6 / 6 |
| Rppg hr | others → buaa-mihr (leave-one-out) | HR MAE ↓ | 2.56 | #1 / 1 |
| Rppg hr | others → mmpd (leave-one-out) | HR MAE ↓ | 10.33 | #1 / 1 |
| Rppg hr | others → pure (leave-one-out) | HR MAE ↓ | 2.86 | #1 / 1 |
| Rppg hr | others → ubfc-rppg (leave-one-out) | HR MAE ↓ | 10.04 | #1 / 1 |
PHASE-Net. (2025). CVPR 2026. https://arxiv.org/abs/2509.24850
@inproceedings{phasenet2025phasenet,
title = {{PHASE-Net}},
year = {2025},
booktitle = {CVPR 2026},
eprint = {2509.24850},
archivePrefix = {arXiv}
}Record updated
