Research map / Video / Video segmentation and tracking
Non-Contrastive Unsupervised Learning of Physiological Signals From Video
Jeremy Speth, Nathan Vance, Patrick Flynn, Adam Czajka
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the bandwidth, sparsity, and variance losses together (ablation) | SiNC · UBFC-rPPG | MAE (bpm): 0.59 ± 0.00 vs 3.08 ± 1.69 (L_b only) | Improved |
| use only sparsity loss L_s (ablation) | SiNC · UBFC-rPPG | MAE (bpm): 0.59 ± 0.00 -> 45.50 ± 1.22 | Hurt |
| use only variance loss L_v (ablation) | SiNC · UBFC-rPPG | MAE (bpm): 0.59 ± 0.00 -> 22.89 ± 2.83 | Hurt |
| omit bandwidth loss L_b (ablation) | SiNC · UBFC-rPPG | MAE (bpm): 0.59 ± 0.00 -> 51.24 ± 5.36 | Hurt |
| omit variance loss L_v (ablation) | SiNC · UBFC-rPPG | MAE (bpm): 0.59 ± 0.00 -> 9.99 ± 2.55 | Hurt |
| omit sparsity loss L_s (ablation) | SiNC · UBFC-rPPG | MAE (bpm): 0.59 ± 0.00 -> 4.18 ± 2.88 | Hurt |
| allow energy in the second harmonic in the sparsity loss (ablation) | SiNC · PURE | MAE (bpm): 0.61 ± 0.06 -> 3.29 ± 1.69 | Hurt |
| vary batch size among 5, 10, 15, and 20 (ablation) | SiNC · PURE | MAE (bpm): 0.73 ± 0.08; 0.65 ± 0.09; 1.75 ± 1.39; 0.61 ± 0.06 | No clear effect |
| remove training augmentations (ablation) | SiNC · PURE | MAE (bpm): 0.61 ± 0.06 -> 33.86 ± 0.83 | Hurt |
| train on unlabeled HKBU-MARs videos rather than rPPG-specific labeled data | SiNC · UBFC-rPPG and PURE cross-dataset testing | MAE (bpm): UBFC: 1.08 ± 0.03; PURE: 2.43 ± 0.20 | Improved |
| train on CelebV-HQ online videos | SiNC · CelebV-HQ training | pulse signal convergence | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Contrast-Phys — Contrast-Phys: Unsupervised Video-Based Remote Physiological Measurement via Spatiotemporal Contrast (ECCV 2022) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The Way to My Heart Is Through Contrastive Learning: Remote Photoplethysmography From Unlabelled Video (ICCV 2021 · similarity 0.92)
- RhythmMamba: Fast, Lightweight, and Accurate Remote Physiological Measurement (AAAI 2025 · similarity 0.89)
- FacePhys: State of the Heart Learning (NeurIPS 2026 · similarity 0.88)
- FactorizePhys: Matrix Factorization for Multidimensional Attention in Remote Physiological Sensing (NeurIPS 2024 · similarity 0.87)
- PhysDiff: Physiology-based Dynamicity Disentangled Diffusion Model for Remote Physiological Measurement (AAAI 2025 · similarity 0.87)
- Remote Photoplethysmography in Real-World and Extreme Lighting Scenarios (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
