Research map / Label-efficient and robust learning / Test-time adaptation
Bi-TTA: Bidirectional Test-Time Adapter for Remote Physiological Measurement
Haodong LI, Hao LU, Yingcong Chen
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 prospective adaptation and retrospective stabilization on top of the two priors (ablation) | Bi-TTA · VIPL-HR | MAE: 7.31 -> 7.09 | Improved |
| add prospective adaptation and retrospective stabilization on top of the two priors (ablation) | Bi-TTA · PURE | MAE: 5.56 -> 5.02 | Improved |
| add prospective adaptation and retrospective stabilization on top of the two priors (ablation) | Bi-TTA · V4V | MAE: 8.97 -> 9.1 | Hurt |
| add retrospective stabilization (ablation) | Bi-TTA · VIPL-HR | MAE: 7.15 -> 7.09 | Improved |
| add prospective adaptation (ablation) | Bi-TTA · PURE | MAE: 5.39 -> 5.02 | Improved |
| add PA and RS on V4V relative to w/o RS variant (ablation) | Bi-TTA · V4V | MAE: 8.93 -> 9.1 | No clear effect |
| use TTA protocol instead of domain generalization | rPPG HR estimation · VIPL-HR | MAE: 7.86 -> 7.09 | Improved |
| use prospective adaptation (sharpness-aware) without retrospective stabilization | Bi-TTA · down-sampled VIPL over many samples | MAE over time | Mixed |
| use priors-only adaptation without PA or RS | Bi-TTA priors · down-sampled VIPL | MAE over time | Hurt |
| use proposed temporal and spatial priors loss instead of ConPhys self-supervised loss (ablation) | Bi-TTA priors · VIPL-HR | MAE: 7.43 -> 7.31 | Improved |
Later papers that compared with this method
1 later accepted paper reports BiTTA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Exposing Mixture and Annotating Confusion for Active Universal Test-Time Adaptation (ICLR 2026 · better on 1 of 1 shared result)
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
- PhysFormer: Facial Video-Based Physiological Measurement With Temporal Difference Transformer (CVPR 2022 · similarity 0.91)
- Remote Photoplethysmography in Real-World and Extreme Lighting Scenarios (CVPR 2025 · similarity 0.91)
- Facial Video-based Remote Physiological Measurement via Self-supervised Learning (TPAMI 2023 · similarity 0.91)
- Continual Learning for Remote Physiological Measurement: Minimize Forgetting and Simplify Inference (ECCV 2024 · similarity 0.91)
- RhythmMamba: Fast, Lightweight, and Accurate Remote Physiological Measurement (AAAI 2025 · similarity 0.90)
- PhysDiff: Physiology-based Dynamicity Disentangled Diffusion Model for Remote Physiological Measurement (AAAI 2025 · similarity 0.89)
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.
