Pith. sign in

REVIEW 4 major objections 6 minor 90 references

Align the GAP: Prior-based Unified Multi-Task Remote Physiological Measurement Framework For Domain Generalization and Personalization

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that one unified framework, GAP, makes multi-task camera-based vital-sign estimation both domain-generalizable and per-person adaptable using the same person-identity representation in opposite directions.

desk verdict A competent extension of PhysMLE that introduces a genuinely new TTPA protocol and a useful driving dataset, but the personalization claim leans on an unvalidated transfer assumption for Z_p. read the letter →

arxiv 2506.16160 v1 pith:TBTCXUND submitted 2025-06-19 cs.CV

Jiyao Wang , Xiao Yang , Hao Lu , Dengbo He , Kaishun Wu This is my paper
classification cs.CV
keywords remotephotoplethysmographymulti-tasklearningdomaingeneralizationtest-timepersonalizedadaptationself-supervisedvitalsignestimationdrivingdatasetindividualbiasmodeling
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper is trying to establish that one multi-task framework, GAP, can solve two problems at once for camera-based vital-sign measurement: generalizing to unseen domains (multi-source synsemantic domain generalization) and then adapting to a particular person at test time (test-time personalized adaptation), using only small configuration changes. The authors argue that the two goals are not opposite if facial videos are disentangled into invariant physiological semantics, per-person bias, and noise, and that the same per-person bias representation can be suppressed during generalization training and amplified during personalization. On six public datasets plus a newly collected driving dataset, they report that the generalization variant GAP-G improves weakly supervised tasks such as SpO2 and RR relative to the previous MSSDG method, and that the personalization variant GAP-P is the only method that improves on every target domain under the TTPA protocol. The paper also acknowledges in its own limitations that GAP-G does not beat the previous method on every task and that real-driving performance remains limited, so the strongest claims are the consistent TTPA gains and the large improvements on weakly labeled tasks.

What carries the argument

The load-bearing mechanism is an auxiliary identity-classification branch that turns the shared representation into a per-person bias vector $Z_p$. During MSSDG training, an orthogonality loss $L_{PE}$ pushes each task-specific representation away from $Z_p$, so the model is forced to rely on invariant physiological semantics; during TTPA, $Z_p$ is injected into the task features by dot-add followed by layer normalization, so the same vector acts as a personalization code. Around this core sit two supporting families: a prior-based augmentation that rescales each channel's mean and standard deviation using the RoR-SpO2 principle, and self-supervised regularizations on the encoder (semantic structure and distribution alignment) and on the decoders (frequency-domain consistency and time-domain inconsistency).

What would settle it

Permute the identity labels during MSSDG training so the auxiliary branch cannot learn true person identity, then run the same GAP-G and GAP-P evaluations on VIPL-HR and HMPC-D; if the SpO2, RR, and per-person adaptation gains remain at or near the reported levels, the identity-bias mechanism is not what carries the result.

Watch

Extended reading notes

Core claim

The central claim is that one architecture with two switchable configurations solves what the authors call multi-source synsemantic domain generalization (MSSDG) and test-time personalized adaptation (TTPA) together. The same trained model, used as GAP-G, learns cross-domain invariant representations by suppressing a person-identity representation $Z_p$ with an orthogonality constraint; used as GAP-P, it fuses $Z_p$ back into task features and adapts to a single individual from unlabeled video. The design is driven by three priors: the Ratio-of-Ratios basis of SpO2 pulse oximetry, which motivates an augmentation that scales channel means and standard deviations without changing physiological state; frequency-domain spatio-temporal consistency of physiological signals; and time-domain inconsistency of BVP morphology across individuals. Empirically the paper reports that GAP-P outperforms all TTA baselines on every target dataset under the TTPA protocol, and that GAP-G substantially improves the weakly labeled tasks, including SpO2 MAE on VIPL-HR dropping from 10.94 (PhysMLE) to 5.06 and RR MAE on HCW dropping from 5.90 to 2.08. The paper also states in its limitations that GAP-G does not beat PhysMLE on every task, notably HR on VIPL-HR and HCW.

Load-bearing premise

The load-bearing assumption is that the identity-classification branch, trained only on identities present in the source domains, learns a per-person bias representation $Z_p$ that transfers to unseen target individuals; if $Z_p$ instead encodes source-domain artifacts, both the MSSDG gains from removing it and the TTPA gains from reinjecting it collapse.

Editorial extensions

If this is right

  • A single deployed model can serve both a population-level generalization mode and a per-user personalization mode, switched only by which losses and fusion operations are active, without any labeled target data.
  • Weakly supervised tasks such as SpO2 and RR benefit most from the prior-based input augmentation, so the approach suggests that label scarcity in rPPG can be attacked at the input level rather than by more complex feature or adapter machinery.
  • The TTPA evaluation protocol, adapting to each new individual chronologically without reusing a sample and without letting users share adaptation, is a usable template for measuring personalization in rPPG and related sensing.
  • If the claimed mechanism holds, the self-supervised losses combining frequency consistency and time-domain inconsistency are sufficient to adapt a multi-task physiological estimator to a single person from unlabeled video.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An implication the paper leaves implicit: the recipe of training a person-identity branch, suppressing it during generalization, and re-injecting it during test-time adaptation is not specific to rPPG; the same switch could apply to other user-specific sensing tasks such as gaze estimation or emotion recognition, where domain-level and person-level shifts coexist.
  • The sensitivity analysis with pi suggests that a single global orthogonality strength is a compromise across tasks, with SpO2 preferring a weaker constraint while HR and RR prefer a stronger one; a per-task pi or a learned schedule would be a direct, testable extension.
  • The paper's own evidence for individual bias rests on a 3-participant breath-hold pilot; a larger independent measurement of RoR-SpO2 intercepts across many participants would either strengthen or weaken the premise that the auxiliary branch models transferable physiology rather than dataset artifacts.
  • The night-time HMPC-D results being better than daytime ones, despite conventional expectations, suggests that variation in illumination, not illumination level, is the harder domain shift; a dataset that records lighting metadata could test this directly.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes GAP, a unified framework for multi-source synsemantic domain generalization (MSSDG) and test-time personalized adaptation (TTPA) in multi-task remote photoplethysmography (rPPG), jointly estimating HR, BVP, SpO2, and RR. The method combines prior-based input augmentation, shared-representation regularization, self-supervised losses based on frequency-domain consistency and time-domain inconsistency, and an auxiliary identity-classification branch whose intermediate representation Z_p is either decorrelated from task features (GAP-G) or fused back for per-person adaptation (GAP-P). Evaluation is performed on six public datasets and a newly collected driving dataset (HMPC-D), reporting improvements over prior MSSDG and TTA baselines, with a substantial reduction in FLOPs and inference time relative to PhysMLE. The paper also defines and benchmarks a TTPA protocol that is new to the multi-task rPPG setting.

Significance. If the claims hold, the paper makes a useful contribution: it provides a unified architecture switchable between generalization and personalization, introduces a new TTPA benchmark, releases a driving dataset with complete physiological labels, and shows a large computational-efficiency gain over PhysMLE (Table 7: 3.37 vs. 34.57 GFLOPs). The TTPA results are directionally consistent across datasets, and the ablation study in Fig. 11 is a valuable check that each loss component contributes. The main risk is that the paper's most novel mechanism, the transferable individual-bias representation Z_p, is asserted rather than demonstrated, and the reported gains are not accompanied by uncertainty estimates. If the transferability assumption is validated and the numerical claims corrected, this would be a solid contribution to the rPPG generalization and personalization literature.

major comments (4)
  1. [Section 3.5.2, Eq. (9), Tables 3-6] The central personalization mechanism rests on an untested assumption: Z_p, learned by classifying N_p+1 source identities with cross-entropy loss L_P, is assumed to encode a transferable individual-bias representation for unseen target individuals. Both the orthogonality removal in L_PE (Eq. 9) and the TTPA dot-add fusion rely on this assumption, but no experiment shows that Z_p generalizes beyond source identities. The 3-participant pilot in Section 3.2.1 demonstrates only that handcrafted RoR intercepts vary across individuals, not that the learned deep representation carries person-specific physiology. Because all TTPA results use the full GAP-P pipeline and the ablation in Fig. 11 does not isolate the Z_p fusion from the shared self-supervised losses, the reported TTPA gains could in principle come from generic adaptation. Please add a control that keeps all TTPA losses but removes Z_p fusion, probe Z_p on unseen identities, and compare it against a representation trained on source artifacts or random labels.
  2. [Section 3.3, Eq. (3)] The component-scaling augmentation, as written, does not preserve the RoR/SpO2 invariant it is claimed to preserve. If the original signal has mean µ and standard deviation σ, the augmented signal X_a = γ*(X_o − µ)/σ + γ*µ has mean γµ and standard deviation γ; hence the AC/DC ratio becomes 1/µ instead of the original σ/µ. The invariance that justifies the augmentation for SpO2 therefore does not hold for Eq. (3). Additionally, the 'PSD unchanged' statement in Fig. 5(d) is at most a shape statement, since the PSD magnitude is scaled by (γ/σ)^2. Please correct the formula or clarify the intended transformation, and verify experimentally that the augmentation preserves the SpO2-relevant semantics.
  3. [Section 4.2.2, Tables 3-6] All results are reported as 5-seed averages without standard deviations, confidence intervals, or significance tests. Several headline margins are small, such as PURE HR MAE 5.27 vs. 5.43 for PhysMLE, V4V HR 7.60 vs. 8.12 for PhysMLE, and HCW RR GAP-P 1.90 vs. 2.00 for AdaODM. The text also misstates several relative improvements: Section 4.3.1 says PURE HR MAE improves by about 10% over PhysMLE, but the table shows a 2.9% improvement; the same paragraph says VIPL-HR SpO2 MAE improves by about 9%, while the table shows a 54% improvement. The claim in Section 4.3.2 that 'GAP-P consistently outperforms all other methods across target domains' is therefore not statistically supported as written. Please report per-seed variability and significance tests, and correct the percentage claims.
  4. [Section 3.6, Eq. (10)] The TTPA optimization objective is stated as LU + LA, but LU includes p3 L_P, the identity cross-entropy loss. In TTPA, target individuals are not in the source-identity output space, which was fixed to N_p+1 identities during MSSDG. The paper does not state whether L_P is disabled, redefined with pseudo-labels, or computed in some other way during TTPA. If L_P is disabled in TTPA, Eq. (10) should indicate this; if it is kept, the training procedure is underspecified. This ambiguity affects the reproducibility of all TTPA experiments, since the loss composition during adaptation is not fully defined.
minor comments (6)
  1. [Section 4.1, Table 7] The text says the parameter count of GAP is 'only two-thirds of PhysMLE,' but Table 7 lists GAP-G with 29.53M parameters and PhysMLE with 24.82M; please reconcile this discrepancy.
  2. [Table 4] The Mixstyle row for HCW RR reports RMSE 3.28, which is implausible given MAE 7.83 and p 0.07 and inconsistent with the other entries in that column; this appears to be a typo for 13.28 or similar.
  3. [Section 3.5.2, Eq. (9)] The expression Z_p · Z_i^T − I is undefined as written if Z_p and Z_i are feature vectors; please clarify whether the outer product, covariance matrix, or a different matrix-based orthogonality penalty is intended.
  4. [Section 4.2.1] The hyperparameter p3 is not listed among the adjusted loss weights (p1, p2, p4-p7); please state its value and whether L_P is active during TTPA.
  5. [Figure 10] The caption refers to 'GAP-A' in subfigure (g), while the text and the rest of the figure use 'GAP-P'; please correct the typo.
  6. [Section 5] The limitation section acknowledges that GAP-G does not improve significantly over PhysMLE in all domains and tasks; consider qualifying the 'consistently outperforms' statements in Sections 4.3.1 and 4.3.2 to match this admission.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular reduction: GAP's gains are measured against external ground truth; the main caveats are reliance on the authors' prior benchmark/pseudo-label scheme and an untested transfer assumption for the identity-bias features, neither of which makes the prediction equal to its input by construction.

full rationale

The central MSSDG and TTPA claims are not circular in the sense targeted by this review. In Tables 3-6, GAP-G and GAP-P are evaluated on held-out target domains with ground-truth HR, SpO2, RR, BVP, and HRV labels from six public datasets plus the new HMPC-D driving dataset, against classical, deep, DG, and TTA baselines. No evaluation metric is identical to any training loss by construction: L_SSA, L_SDA, L_FC, L_TIC, and L_PE are self-supervised or orthogonality losses, while the reported MAE/RMSE/p values are computed with external labels. The RoR-based component scaling in Eq. (3) preserves the ratio-of-ratios by construction, but it is used as a data-augmentation prior for self-supervision, not as a fitted 'prediction' of SpO2; the SpO2 and RR improvements on target domains are measured against ground truth. The personalization mechanism in Section 3.5.2 does rest on an assumption that the identity-classification representation Z_p transfers to unseen individuals, and the paper itself cautions in Section 4.5.3 that 'we cannot determine that this performance change is brought about by individual heterogeneity' in real driving environments; this is a validity or assumption risk, not a circular derivation. The framework does lean on the authors' own prior work (PhysMLE, J. Wang et al., 2025) for the MSSDG benchmark, pseudo-label construction, and feature-gate details, and the TTPA protocol is newly defined and instantiated in this paper, so independent confirmation is somewhat reduced. Nevertheless, the empirical comparisons and ablations (Figure 11) give the central claims content outside of any fitted parameter or self-citation chain. Therefore, no concrete circular step is exhibited; the score is 2 for minor self-citation weight rather than for a reduction of the claimed predictions to their inputs.

Assumptions & free parameters 4 free parameters · 6 assumptions · 1 invented entities

GAP rests on four tuned quantities (pi, loss weights p1-p7, augmentation range gamma, offset range Delta_T), five domain assumptions inherited from physiology and prior rPPG literature (frequency-domain consistency, RoR invariance, time-domain BVP inconsistency as person-specific, pseudo-label inter-task correlations, and the Bayesian independence split of Eq. 1), and one paper-specific assumption: that the identity-classification branch trained on source identities produces a transferable individual-bias representation. The invented-entity count is low: only Z_p qualifies, and it lacks an external handle, although the ablation gives internal evidence for its role.

free parameters (4)
  • orthogonality threshold pi in L_PE = 0.1
    Eq. (9); chosen via validation performance. Sensitivity analysis (Fig. 9) shows HR/RR prefer pi=0.3 and SpO2 prefers 0.0; the authors acknowledge the global choice is a compromise.
  • self-supervised loss weights p1..p7 = p1=0.0001, p2=0.001, p4-p7=0.01
    Eq. (10); set by 'aligning with the numeric scale of each loss and their convergence across source domains', i.e., tuned on validation data.
  • augmentation scaling range gamma = [0.8, 2.2]
    Prior-based augmentation, Eq. (3); range taken directly from cited prior work (S. Chen et al. 2023), not derived in this paper.
  • temporal offset range Delta_T = uniform over 0 to 30 frames
    Augmentation offset simulating 0 to 1 second of sampling delay; chosen by hand, no justification beyond the physical interpretation.
assumptions (6)
  • domain assumption Bayesian factorization of P(X_phys, X_noise | Y) with X_phys and X_noise independent (Eq. 1)
    Motivates separating semantic from noise/biased components; independence is asserted, citing prior work (Y. Zhang et al. 2024).
  • domain assumption Frequency-domain spatio-temporal consistency of HR/RR/SpO2 signals
    Basis of L_FC; adopted from prior literature (Z. Sun and Li 2024; H. Li et al. 2024).
  • domain assumption Time-domain BVP inconsistency is person-specific and informative
    Basis of L_TIC; supported by physiology citations (Lister et al. 2012; Persson 1996) and the authors' Fig. 4 observation using the GREEN algorithm.
  • domain assumption Scaling each STMap row and channel by gamma preserves task semantics via RoR invariance
    Basis of the prior-based augmentation, Eq. (3); RoR invariance from Guazzi et al. 2015, and PSD invariance argued for BVP in Fig. 5(d).
  • ad hoc to paper The person-classification auxiliary task produces a transferable individual-bias representation Z_p
    Section 3.5.2; the classifier output space is fixed to N_p+1 source identities while TTPA adapts to unseen individuals, so transferability of Z_p is assumed rather than demonstrated.
  • domain assumption Inter-task pseudo-labels (e.g., RR extracted from BVP frequency peaks) provide valid soft supervision
    Section 3.5.1; inherited from the authors' prior PhysMLE work (J. Wang et al. 2025), with soft constraints applied because the correlations are unstable.
invented entities (1)
  • Individual-bias representation Z_p
    purpose: Explicitly models person-specific physiological mapping for removal during MSSDG and re-injection during TTPA
    Learned via an identity-classification auxiliary task; there is no external falsifiable handle (e.g., a predicted measurable quantity). The ablation in Fig. 11 gives only internal evidence, so transferability to unseen identities remains unverified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Align the GAP: Prior-based Unified Multi-Task Remote Physiological Measurement Framework For Domain Generalization and Personalization." pith.science (2026). https://pith.science/paper/TBTCXUND

@misc{pith2026250616160,
  author       = {Pith},
  title        = {Pith review of: Align the GAP: Prior-based Unified Multi-Task Remote Physiological Measurement Framework For Domain Generalization and Personalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TBTCXUND}},
  note         = {Machine review of arXiv:2506.16160}
}
read the original abstract

Multi-source synsemantic domain generalization (MSSDG) for multi-task remote physiological measurement seeks to enhance the generalizability of these metrics and attracts increasing attention. However, challenges like partial labeling and environmental noise may disrupt task-specific accuracy. Meanwhile, given that real-time adaptation is necessary for personalized products, the test-time personalized adaptation (TTPA) after MSSDG is also worth exploring, while the gap between previous generalization and personalization methods is significant and hard to fuse. Thus, we proposed a unified framework for MSSD\textbf{G} and TTP\textbf{A} employing \textbf{P}riors (\textbf{GAP}) in biometrics and remote photoplethysmography (rPPG). We first disentangled information from face videos into invariant semantics, individual bias, and noise. Then, multiple modules incorporating priors and our observations were applied in different stages and for different facial information. Then, based on the different principles of achieving generalization and personalization, our framework could simultaneously address MSSDG and TTPA under multi-task remote physiological estimation with minimal adjustments. We expanded the MSSDG benchmark to the TTPA protocol on six publicly available datasets and introduced a new real-world driving dataset with complete labeling. Extensive experiments that validated our approach, and the codes along with the new dataset will be released.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

90 extracted references · 71 canonical work pages

  1. [1]

    write newline

    " write newline " cite write " FUNCTION editor.postfix editor num.names #1 > "( )" "( )" if FUNCTION editor.trans.postfix editor num.names #1 > "( )" "( )" if FUNCTION trans.postfix translator num.names #1 > "( )" "( )" if FUNCTION authors.editors.reflist.apa5 'field := 'dot := field num.names 'numnames := numnames 'format.num.names := format.num.names na...

  2. [2]

    , Yang, G Y

    ahn2024style APACrefauthors Ahn, W J. , Yang, G Y. , Choi, H D. Lim, M T. APACrefauthors \ 2024 . Style Blind Domain Generalized Semantic Segmentation via Covariance Alignment and Semantic Consistence Contrastive Learning Style blind domain generalized semantic segmentation via covariance alignment and semantic consistence contrastive learning . Proceedin...

    2024

  3. [3]

    , Onishi, Y

    akamatsu2023blood APACrefauthors Akamatsu, Y. , Onishi, Y. Imaoka, H. APACrefauthors \ 2023 . Blood oxygen saturation estimation from facial video via dc and ac components of spatio-temporal map Blood oxygen saturation estimation from facial video via dc and ac components of spatio-temporal map . ICASSP 2023-2023 IEEE International Conference on Acoustics...

    2023

  4. [4]

    , Hahn, O

    bahmani2022semantic APACrefauthors Bahmani, S. , Hahn, O. , Zamfir, E. , Araslanov, N. , Cremers, D. Roth, S. APACrefauthors \ 2022 . Semantic Self-adaptation: Enhancing Generalization with a Single Sample Semantic self-adaptation: Enhancing generalization with a single sample . arXiv preprint arXiv:2208.05788 ,

    arXiv 2022

  5. [5]

    APACrefauthors \ 2015

    bal2015non APACrefauthors Bal, U. APACrefauthors \ 2015 . Non-contact estimation of heart rate and oxygen saturation using ambient light Non-contact estimation of heart rate and oxygen saturation using ambient light . Biomedical optics express 6 1 86--97,

    2015

  6. [6]

    , Thomas Bigger Jr, J

    berntson1997heart APACrefauthors Berntson, G.G. , Thomas Bigger Jr, J. , Eckberg, D.L. , Grossman, P. , Kaufmann, P.G. , Malik, M. others APACrefauthors \ 1997 . Heart rate variability: origins, methods, and interpretive caveats Heart rate variability: origins, methods, and interpretive caveats . Psychophysiology 34 6 623--648,

    1997

  7. [7]

    , Macwan, R

    bobbia2019unsupervised APACrefauthors Bobbia, S. , Macwan, R. , Benezeth, Y. , Mansouri, A. Dubois, J. APACrefauthors \ 2019 . Unsupervised skin tissue segmentation for remote photoplethysmography Unsupervised skin tissue segmentation for remote photoplethysmography . Pattern Recognition Letters 124 82--90,

    2019

  8. [8]

    , McDuff, D

    braun2024suboptimal APACrefauthors Braun, B. , McDuff, D. Holz, C. APACrefauthors \ 2024 . How Suboptimal is Training rPPG Models with Videos and Targets from Different Body Sites? How suboptimal is training rppg models with videos and targets from different body sites? Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proc...

    2024

Show all 90 references
  1. [9]

    , Zhang, C

    chang2023pepnet APACrefauthors Chang, J. , Zhang, C. , Hui, Y. , Leng, D. , Niu, Y. , Song, Y. Gai, K. APACrefauthors \ 2023 . Pepnet: Parameter and embedding personalized network for infusing with personalized prior information Pepnet: Parameter and embedding personalized net...

    2023

  2. [10]

    , Ho, S.K

    chen2023deep APACrefauthors Chen, S. , Ho, S.K. , Chin, J.W. , Luo, K.H. , Chan, T.T. , So, R.H. Wong, K.L. APACrefauthors \ 2023 . Deep learning-based image enhancement for robust remote photoplethysmography in various illumination scenarios Deep learning-based image enhancem...

    2023

  3. [11]

    \ McDuff, D

    chen2018deepphys APACrefauthors Chen, W. \ McDuff, D. APACrefauthors \ 2018 . Deepphys: Video-based physiological measurement using convolutional attention networks Deepphys: Video-based physiological measurement using convolutional attention networks . Proceedings of the euro...

    2018

  4. [12]

    , Liu, R

    cheng2023motion APACrefauthors Cheng, J. , Liu, R. , Li, J. , Song, R. , Liu, Y. Chen, X. APACrefauthors \ 2023 . Motion-robust respiratory rate estimation from camera videos via fusing pixel movement and pixel intensity information Motion-robust respiratory rate estimation fr...

    2023

  5. [13]

    , Wang, Y

    chi2021test APACrefauthors Chi, Z. , Wang, Y. , Yu, Y. Tang, J. APACrefauthors \ 2021 . Test-time fast adaptation for dynamic scene deblurring via meta-auxiliary learning Test-time fast adaptation for dynamic scene deblurring via meta-auxiliary learning . Proceedings of the IE...

    2021

  6. [14]

    das2021bvpnet APACrefauthors Das, A. , Lu, H. , Han, H. , Dantcheva, A. , Shan, S. Chen, X. APACrefauthors \ 2021 . BVPNet: Video-to-BVP Signal Prediction for Remote Heart Rate Estimation Bvpnet: Video-to-bvp signal prediction for remote heart rate estimation . 2021 16th IEEE ...

    2021

  7. [15]

    \ Jeanne, V

    de2013robust APACrefauthors De Haan, G. \ Jeanne, V. APACrefauthors \ 2013 . Robust pulse rate from chrominance-based rPPG Robust pulse rate from chrominance-based rppg . IEEE Transactions on Biomedical Engineering 60 10 2878--2886,

    2013

  8. [16]

    , Liu, S Q

    du2021weakly APACrefauthors Du, J. , Liu, S Q. , Zhang, B. Yuen, P.C. APACrefauthors \ 2021 . Weakly supervised rppg estimation for respiratory rate estimation Weakly supervised rppg estimation for respiratory rate estimation . Proceedings of the IEEE/CVF International Confere...

    2021

  9. [17]

    , Liu, S Q

    du2023dual APACrefauthors Du, J. , Liu, S Q. , Zhang, B. Yuen, P.C. APACrefauthors \ 2023 . Dual-bridging with adversarial noise generation for domain adaptive rppg estimation Dual-bridging with adversarial noise generation for domain adaptive rppg estimation . Proceedings of ...

    2023

  10. [18]

    , Dellandr \'e a, E

    duboudin2022learning APACrefauthors Duboudin, T. , Dellandr \'e a, E. , Abgrall, C. , H \'e naff, G. Chen, L. APACrefauthors \ 2022 . Learning Less Generalizable Patterns with an Asymmetrically Trained Double Classifier for Better Test-Time Adaptation Learning less generalizab...

    2022 arXiv

  11. [19]

    , Abbeel, P

    finn2017model APACrefauthors Finn, C. , Abbeel, P. Levine, S. APACrefauthors \ 2017 . Model-agnostic meta-learning for fast adaptation of deep networks Model-agnostic meta-learning for fast adaptation of deep networks . International conference on machine learning Internationa...

    2017

  12. [20]

    \ Lempitsky, V

    ganin2015unsupervised APACrefauthors Ganin, Y. \ Lempitsky, V. APACrefauthors \ 2015 . Unsupervised domain adaptation by backpropagation Unsupervised domain adaptation by backpropagation . International conference on machine learning International conference on machine learnin...

    2015

  13. [21]

    , Ghanadian, H

    ghodratigohar2019remote APACrefauthors Ghodratigohar, M. , Ghanadian, H. Al Osman, H. APACrefauthors \ 2019 . A remote respiration rate measurement method for non-stationary subjects using CEEMDAN and machine learning A remote respiration rate measurement method for non-statio...

    2019

  14. [22]

    , Villarroel, M

    guazzi2015non APACrefauthors Guazzi, A.R. , Villarroel, M. , Jorge, J. , Daly, J. , Frise, M.C. , Robbins, P.A. Tarassenko, L. APACrefauthors \ 2015 . Non-contact measurement of oxygen saturation with an RGB camera Non-contact measurement of oxygen saturation with an rgb camer...

    2015

  15. [23]

    , Zhang, J

    han2020leveraging APACrefauthors Han, D. , Zhang, J. Shan, S. APACrefauthors \ 2020 . Leveraging auxiliary tasks for height and weight estimation by multi task learning Leveraging auxiliary tasks for height and weight estimation by multi task learning . 2020 IEEE International...

    2020

  16. [24]

    , Zhang, X

    he2016deep APACrefauthors He, K. , Zhang, X. , Ren, S. Sun, J. APACrefauthors \ 2016 . Deep residual learning for image recognition Deep residual learning for image recognition . Proceedings of the IEEE conference on computer vision and pattern recognition Proceedings of the i...

    2016

  17. [25]

    , Abboud, F.M

    heistad1973interaction APACrefauthors Heistad, D.D. , Abboud, F.M. , Mark, A.L. Schmid, P.G. APACrefauthors \ 1973 . Interaction of thermal and baroreceptor reflexes in man. Interaction of thermal and baroreceptor reflexes in man. Journal of applied physiology 35 5 581--586,

    1973

  18. [26]

    hu2023contactless APACrefauthors Hu, M. , Wu, X. , Wang, X. , Xing, Y. , An, N. Shi, P. APACrefauthors \ 2023 . Contactless blood oxygen estimation from face videos: A multi-model fusion method based on deep learning Contactless blood oxygen estimation from face videos: A mult...

    2023

  19. [27]

    , Chen, T H

    huang2024fully APACrefauthors Huang, P K. , Chen, T H. , Chan, Y T. , Chen, K W. Hsu, C T. APACrefauthors \ 2024 . Fully Test-Time rPPG Estimation via Synthetic Signal-Guided Feature Learning Fully test-time rppg estimation via synthetic signal-guided feature learning . arXiv ...

    2024 arXiv

  20. [28]

    \ Bartsch, R.P

    ivanov2014network APACrefauthors Ivanov, P.C. \ Bartsch, R.P. APACrefauthors \ 2014 . Network physiology: mapping interactions between networks of physiologic networks Network physiology: mapping interactions between networks of physiologic networks . Networks of Networks: the...

    2014

  21. [29]

    , Wang, W

    janssen2015video APACrefauthors Janssen, R. , Wang, W. , Mo c o, A. De Haan, G. APACrefauthors \ 2015 . Video-based respiration monitoring with automatic region of interest detection Video-based respiration monitoring with automatic region of interest detection . Physiological...

    2015

  22. [30]

    , Hao, L

    jiyao2023DGrppg APACrefauthors Jiyao, W. , Hao, L. , Hu, H. , Yingcong, C. , Dengbo, H. Kaishun, W. APACrefauthors \ 2024 . Generalizable Remote Physiological Measurement via Semantic-Sheltered Alignment and Plausible Style Randomization Generalizable remote physiological meas...

    2024

  23. [31]

    APACrefauthors \ 1986

    johnson1986nonthermoregulatory APACrefauthors Johnson, J. APACrefauthors \ 1986 . Nonthermoregulatory control of human skin blood flow Nonthermoregulatory control of human skin blood flow . Journal of Applied Physiology 61 5 1613--1622,

    1986

  24. [32]

    , Ayache, M

    klingner2022continual APACrefauthors Klingner, M. , Ayache, M. Fingscheidt, T. APACrefauthors \ 2022 . Continual BatchNorm Adaptation (CBNA) for Semantic Segmentation Continual batchnorm adaptation (cbna) for semantic segmentation . IEEE Transactions on Intelligent Transportat...

    2022

  25. [33]

    , Zhao, Y

    kong2013non APACrefauthors Kong, L. , Zhao, Y. , Dong, L. , Jian, Y. , Jin, X. , Li, B. Wu, H. APACrefauthors \ 2013 . Non-contact detection of oxygen saturation based on visible light imaging device using ambient light Non-contact detection of oxygen saturation based on visib...

    2013

  26. [34]

    , Caballero, E

    krueger2021out APACrefauthors Krueger, D. , Caballero, E. , Jacobsen, J H. , Zhang, A. , Binas, J. , Zhang, D. Courville, A. APACrefauthors \ 2021 . Out-of-distribution generalization via risk extrapolation (rex) Out-of-distribution generalization via risk extrapolation (rex) ...

    2021

  27. [35]

    li2024bi APACrefauthors Li, H. , Lu, H. Chen, Y C. APACrefauthors \ 2024 . Bi-TTA: Bidirectional Test-Time Adapter for Remote Physiological Measurement Bi-tta: Bidirectional test-time adapter for remote physiological measurement . European Conference on Computer Vision Europea...

    2024

  28. [36]

    , Hao, M

    li2021test APACrefauthors Li, Y. , Hao, M. , Di, Z. , Gundavarapu, N.B. Wang, X. APACrefauthors \ 2021 . Test-time personalization with a transformer for human pose estimation Test-time personalization with a transformer for human pose estimation . Advances in Neural Informati...

    2021

  29. [37]

    liang2020we APACrefauthors Liang, J. , Hu, D. Feng, J. APACrefauthors \ 2020 . Do we really need to access the source data? source hypothesis transfer for unsupervised domain adaptation Do we really need to access the source data? source hypothesis transfer for unsupervised do...

    2020

  30. [38]

    , Wright, P.A

    lister2012optical APACrefauthors Lister, T. , Wright, P.A. Chappell, P.H. APACrefauthors \ 2012 . Optical properties of human skin Optical properties of human skin . Journal of biomedical optics 17 9 090901--090901,

    2012

  31. [39]

    liu2022towards APACrefauthors Liu, H. , Wu, Z. , Li, L. , Salehkalaibar, S. , Chen, J. Wang, K. APACrefauthors \ 2022 . Towards multi-domain single image dehazing via test-time training Towards multi-domain single image dehazing via test-time training . Proceedings of the IEEE...

    2022

  32. [40]

    , Fromm, J

    liu2020multi APACrefauthors Liu, X. , Fromm, J. , Patel, S. McDuff, D. APACrefauthors \ 2020 . Multi-task temporal shift attention networks for on-device contactless vitals measurement Multi-task temporal shift attention networks for on-device contactless vitals measurement . ...

    2020

  33. [41]

    , Hill, B

    liu2023efficientphys APACrefauthors Liu, X. , Hill, B. , Jiang, Z. , Patel, S. McDuff, D. APACrefauthors \ 2023 . Efficientphys: Enabling simple, fast and accurate camera-based cardiac measurement Efficientphys: Enabling simple, fast and accurate camera-based cardiac measureme...

    2023

  34. [42]

    , Kothari, P

    liu2021ttt++ APACrefauthors Liu, Y. , Kothari, P. , Van Delft, B. , Bellot-Gurlet, B. , Mordan, T. Alahi, A. APACrefauthors \ 2021 . Ttt++: When does self-supervised test-time training fail or thrive? Ttt++: When does self-supervised test-time training fail or thrive? Advances...

    2021

  35. [43]

    , Han, H

    lu2021dual APACrefauthors Lu, H. , Han, H. Zhou, S.K. APACrefauthors \ 2021 . Dual-gan: Joint bvp and noise modeling for remote physiological measurement Dual-gan: Joint bvp and noise modeling for remote physiological measurement . Proceedings of the IEEE/CVF Conference on Com...

    2021

  36. [44]

    , Niu, X

    lu2024gpt APACrefauthors Lu, H. , Niu, X. , Wang, J. , Wang, Y. , Hu, Q. , Tang, J. others APACrefauthors \ 2024 . GPT as Psychologist? Preliminary Evaluations for GPT-4V on Visual Affective Computing Gpt as psychologist? preliminary evaluations for gpt-4v on visual affective ...

    2024

  37. [45]

    lu2023neuron APACrefauthors Lu, H. , Yu, Z. , Niu, X. Chen, Y C. APACrefauthors \ 2023 . Neuron structure modeling for generalizable remote physiological measurement Neuron structure modeling for generalizable remote physiological measurement . Proceedings of the IEEE/CVF Conf...

    2023

  38. [46]

    , Liang, J

    lv2022causality APACrefauthors Lv, F. , Liang, J. , Li, S. , Zang, B. , Liu, C.H. , Wang, Z. Liu, D. APACrefauthors \ 2022 . Causality inspired representation learning for domain generalization Causality inspired representation learning for domain generalization . Proceedings ...

    2022

  39. [47]

    , Kim, T

    min2023meta APACrefauthors Min, C. , Kim, T. Lim, J. APACrefauthors \ 2023 . Meta-Learning for Adaptation of Deep Optical Flow Networks Meta-learning for adaptation of deep optical flow networks . Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision...

    2023

  40. [48]

    , Padhy, S

    nado2020evaluating APACrefauthors Nado, Z. , Padhy, S. , Sculley, D. , D'Amour, A. , Lakshminarayanan, B. Snoek, J. APACrefauthors \ 2020 . Evaluating prediction-time batch normalization for robustness under covariate shift Evaluating prediction-time batch normalization for ro...

    2020 arXiv

  41. [49]

    , Liu, Y

    narayanswamy2024bigsmall APACrefauthors Narayanswamy, G. , Liu, Y. , Yang, Y. , Ma, C. , Liu, X. , McDuff, D. Patel, S. APACrefauthors \ 2024 . Bigsmall: Efficient multi-task learning for disparate spatial and temporal physiological measurements Bigsmall: Efficient multi-task ...

    2024

  42. [50]

    niu2022efficient APACrefauthors Niu, S. , Wu, J. , Zhang, Y. , Chen, Y. , Zheng, S. , Zhao, P. Tan, M. APACrefauthors \ 2022 . Efficient test-time model adaptation without forgetting Efficient test-time model adaptation without forgetting . International conference on machine ...

    2022

  43. [51]

    niu2023towards APACrefauthors Niu, S. , Wu, J. , Zhang, Y. , Wen, Z. , Chen, Y. , Zhao, P. Tan, M. APACrefauthors \ 2023 . Towards stable test-time adaptation in dynamic wild world Towards stable test-time adaptation in dynamic wild world . arXiv preprint arXiv:2302.12400 ,

    2023 arXiv

  44. [52]

    , Shan, S

    niu2019rhythmnet APACrefauthors Niu, X. , Shan, S. , Han, H. Chen, X. APACrefauthors \ 2019 . Rhythmnet: End-to-end heart rate estimation from face via spatial-temporal representation Rhythmnet: End-to-end heart rate estimation from face via spatial-temporal representation . I...

    2019

  45. [53]

    , Marks, T.K

    nowara2020near APACrefauthors Nowara, E.M. , Marks, T.K. , Mansour, H. Veeraraghavan, A. APACrefauthors \ 2020 . Near-infrared imaging photoplethysmography during driving Near-infrared imaging photoplethysmography during driving . IEEE transactions on intelligent transportatio...

    2020

  46. [54]

    , Bonnici, T

    Orphanidou2015SignalQualityIF APACrefauthors Orphanidou, C. , Bonnici, T. , Charlton, P.H. , Clifton, D.A. , Vallance, D. Tarassenko, L. APACrefauthors \ 2015 . Signal-Quality Indices for the Electrocardiogram and Photoplethysmogram: Derivation and Applications to Wireless Mon...

    2015

  47. [55]

    , Neitz, A

    parascandolo2020learning APACrefauthors Parascandolo, G. , Neitz, A. , Orvieto, A. , Gresele, L. Sch \"o lkopf, B. APACrefauthors \ 2020 . Learning explanations that are hard to vary Learning explanations that are hard to vary . arXiv preprint arXiv:2009.00329 ,

    2020 arXiv

  48. [56]

    , Harvey, R

    peper2007there APACrefauthors Peper, E. , Harvey, R. , Lin, I M. , Tylova, H. Moss, D. APACrefauthors \ 2007 . Is there more to blood volume pulse than heart rate variability, respiratory sinus arrhythmia, and cardiorespiratory synchrony? Is there more to blood volume pulse th...

    2007

  49. [57]

    APACrefauthors \ 1996

    persson1996modulation APACrefauthors Persson, P.B. APACrefauthors \ 1996 . Modulation of cardiovascular control mechanisms and their interaction Modulation of cardiovascular control mechanisms and their interaction . Physiological reviews 76 1 193--244,

    1996

  50. [58]

    revanur2021first APACrefauthors Revanur, A. , Li, Z. , Ciftci, U.A. , Yin, L. Jeni, L.A. APACrefauthors \ 2021 . The first vision for vitals (v4v) challenge for non-contact video-based physiological estimation The first vision for vitals (v4v) challenge for non-contact video-b...

    2021

  51. [59]

    , Rusak, E

    schneider2020improving APACrefauthors Schneider, S. , Rusak, E. , Eck, L. , Bringmann, O. , Brendel, W. Bethge, M. APACrefauthors \ 2020 . Improving robustness against common corruptions by covariate shift adaptation Improving robustness against common corruptions by covariate...

    2020

  52. [60]

    shi2020towards APACrefauthors Shi, Y. , Yu, X. , Sohn, K. , Chandraker, M. Jain, A.K. APACrefauthors \ 2020 . Towards universal representation learning for deep face recognition Towards universal representation learning for deep face recognition . Proceedings of the IEEE/CVF C...

    2020

  53. [61]

    , Argod, J

    smith1999pulse APACrefauthors Smith, R.P. , Argod, J. , P \'e pin, J L. L \'e vy, P.A. APACrefauthors \ 1999 . Pulse transit time: an appraisal of potential clinical applications Pulse transit time: an appraisal of potential clinical applications . Thorax 54 5 452--457,

    1999

  54. [62]

    , Cho, S

    soh2020meta APACrefauthors Soh, J.W. , Cho, S. Cho, N.I. APACrefauthors \ 2020 . Meta-transfer learning for zero-shot super-resolution Meta-transfer learning for zero-shot super-resolution . Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Proc...

    2020

  55. [63]

    , M \"u ller, S

    stricker2014non APACrefauthors Stricker, R. , M \"u ller, S. Gross, H M. APACrefauthors \ 2014 . Non-contact video-based pulse rate measurement on a mobile service robot Non-contact video-based pulse rate measurement on a mobile service robot . The 23rd IEEE International Symp...

    2014

  56. [64]

    sun2024defense APACrefauthors Sun, H. , Fu, L. , Li, J. , Guo, Q. , Meng, Z. , Zhang, T. Yu, H. APACrefauthors \ 2024 . Defense against adversarial cloud attack on remote sensing salient object detection Defense against adversarial cloud attack on remote sensing salient object...

    2024

  57. [65]

    , Zhang, X

    10.1145/3581783.3612265 APACrefauthors Sun, W. , Zhang, X. , Lu, H. , Chen, Y. , Ge, Y. , Huang, X. Chen, Y. APACrefauthors \ 2023 1 . Resolve Domain Conflicts for Generalizable Remote Physiological Measurement Resolve domain conflicts for generalizable remote physiological me...

    2023

  58. [66]

    , Zhang, X

    sun2023resolve APACrefauthors Sun, W. , Zhang, X. , Lu, H. , Chen, Y. , Ge, Y. , Huang, X. Chen, Y. APACrefauthors \ 2023 2 . Resolve domain conflicts for generalizable remote physiological measurement Resolve domain conflicts for generalizable remote physiological measurement...

    2023

  59. [67]

    sun2024 APACrefauthors Sun, Z. \ Li, X. APACrefauthors \ 2024 . Contrast-Phys+: Unsupervised and Weakly-supervised Video-based Remote Physiological Measurement via Spatiotemporal Contrast Contrast-phys+: Unsupervised and weakly-supervised video-based remote physiological measu...

    2024

  60. [68]

    , Villarroel, M

    tarassenko2014non APACrefauthors Tarassenko, L. , Villarroel, M. , Guazzi, A. , Jorge, J. , Clifton, D. Pugh, C. APACrefauthors \ 2014 . Non-contact video-based vital sign monitoring using ambient light and auto-regressive models Non-contact video-based vital sign monitoring u...

    2014

  61. [69]

    , Svaasand, L.O

    verkruysse2008remote APACrefauthors Verkruysse, W. , Svaasand, L.O. Nelson, J.S. APACrefauthors \ 2008 . Remote plethysmographic imaging using ambient light. Remote plethysmographic imaging using ambient light. Optics express 16 26 21434--21445,

    2008

  62. [70]

    , Huang, C

    wang2024association APACrefauthors Wang, A. , Huang, C. , Wang, J. He, D. APACrefauthors \ 2024 . The association between physiological and eye-tracking metrics and cognitive load in drivers: A meta-analysis The association between physiological and eye-tracking metrics and co...

    2024

  63. [71]

    , Shelhamer, E

    wang2020tent APACrefauthors Wang, D. , Shelhamer, E. , Liu, S. , Olshausen, B. Darrell, T. APACrefauthors \ 2020 . Tent: Fully test-time adaptation by entropy minimization Tent: Fully test-time adaptation by entropy minimization . arXiv preprint arXiv:2006.10726 ,

    2020 arXiv

  64. [72]

    , Huang, C

    wang2024rethink APACrefauthors Wang, J. , Huang, C. , Xie, W. , He, D. Tu, R. APACrefauthors \ 2024 . Rethink data-driven human behavior prediction: A Psychology-powered Explainable Neural Network Rethink data-driven human behavior prediction: A psychology-powered explainable ...

    2024

  65. [73]

    wang2024generalizable APACrefauthors Wang, J. , Lu, H. , Han, H. , Chen, Y. , He, D. Wu, K. APACrefauthors \ 2024 . Generalizable Remote Physiological Measurement via Semantic-Sheltered Alignment and Plausible Style Randomization Generalizable remote physiological measurement ...

    2024

  66. [74]

    wang2023hierarchical APACrefauthors Wang, J. , Lu, H. , Wang, A. , Chen, Y. He, D. APACrefauthors \ 2024 . Hierarchical Style-Aware Domain Generalization for Remote Physiological Measurement Hierarchical style-aware domain generalization for remote physiological measurement . ...

    2024

  67. [75]

    wang2024physmle APACrefauthors Wang, J. , Lu, H. , Wang, A. , Yang, X. , Chen, Y. , He, D. Wu, K. APACrefauthors \ 2025 . PhysMLE: Generalizable and Priors-Inclusive Multi-task Remote Physiological Measurement Physmle: Generalizable and priors-inclusive multi-task remote physi...

    2025

  68. [76]

    , Yang, X

    wang2024efficient APACrefauthors Wang, J. , Yang, X. , Wang, Z. , Wei, X. , Wang, A. , He, D. Wu, K. APACrefauthors \ 2024 . Efficient Mixture-of-Expert for Video-based Driver State and Physiological Multi-task Estimation in Conditional Autonomous Driving Efficient mixture-of-...

    2024

  69. [77]

    \ Deng, W

    wang2018deep APACrefauthors Wang, M. \ Deng, W. APACrefauthors \ 2018 . Deep visual domain adaptation: A survey Deep visual domain adaptation: A survey . Neurocomputing 312 135--153,

    2018

  70. [78]

    , Den Brinker, A.C

    wang2016algorithmic APACrefauthors Wang, W. , Den Brinker, A.C. , Stuijk, S. De Haan, G. APACrefauthors \ 2016 . Algorithmic principles of remote PPG Algorithmic principles of remote ppg . IEEE Transactions on Biomedical Engineering 64 7 1479--1491,

    2016

  71. [79]

    , Tsai, Y H

    wang2022semi APACrefauthors Wang, Y. , Tsai, Y H. , Hung, W C. , Ding, W. , Liu, S. Yang, M H. APACrefauthors \ 2022 . Semi-supervised multi-task learning for semantics and depth Semi-supervised multi-task learning for semantics and depth . Proceedings of the IEEE/CVF winter c...

    2022

  72. [80]

    , CHen, G

    10552776 APACrefauthors Wu, Y. , CHen, G. , Ye, L. , Jia, Y. , Liu, Z. Wang, Y. APACrefauthors \ 2024 . TTAGaze: Self-Supervised Test-Time Adaptation for Personalized Gaze Estimation Ttagaze: Self-supervised test-time adaptation for personalized gaze estimation . IEEE Transact...

    2024

  73. [81]

    , Chen, W

    xi2020image APACrefauthors Xi, L. , Chen, W. , Zhao, C. , Wu, X. Wang, J. APACrefauthors \ 2020 . Image enhancement for remote photoplethysmography in a low-light environment Image enhancement for remote photoplethysmography in a low-light environment . 2020 15th IEEE Internat...

    2020

  74. [82]

    xie2024sfda APACrefauthors Xie, Y. , Yu, Z. , Wu, B. , Xie, W. Shen, L. APACrefauthors \ 2024 . SFDA-rPPG: Source-Free Domain Adaptive Remote Physiological Measurement with Spatio-Temporal Consistency Sfda-rppg: Source-free domain adaptive remote physiological measurement with...

    2024 arXiv

  75. [83]

    , Wang, Y

    yang2021generalized APACrefauthors Yang, S. , Wang, Y. , Van De Weijer, J. , Herranz, L. Jui, S. APACrefauthors \ 2021 . Generalized source-free domain adaptation Generalized source-free domain adaptation . Proceedings of the IEEE/CVF international conference on computer visio...

    2021

  76. [84]

    , Peng, W

    yu2019remote APACrefauthors Yu, Z. , Peng, W. , Li, X. , Hong, X. Zhao, G. APACrefauthors \ 2019 . Remote heart rate measurement from highly compressed facial videos: an end-to-end deep learning solution with video enhancement Remote heart rate measurement from highly compress...

    2019

  77. [85]

    , Shen, Y

    yu2023physformer++ APACrefauthors Yu, Z. , Shen, Y. , Shi, J. , Zhao, H. , Cui, Y. , Zhang, J. Zhao, G. APACrefauthors \ 2023 . Physformer++: Facial video-based physiological measurement with slowfast temporal difference transformer Physformer++: Facial video-based physiologic...

    2023

  78. [86]

    , Nie, X

    zhang2020inference APACrefauthors Zhang, J. , Nie, X. Feng, J. APACrefauthors \ 2020 . Inference stage optimization for cross-scenario 3d human pose estimation Inference stage optimization for cross-scenario 3d human pose estimation . Advances in Neural Information Processing ...

    2020

  79. [87]

    \ Chen, Y C

    zhang2023adaptive APACrefauthors Zhang, X. \ Chen, Y C. APACrefauthors \ 2023 . Adaptive domain generalization via online disagreement minimization Adaptive domain generalization via online disagreement minimization . IEEE Transactions on Image Processing ,

    2023

  80. [88]

    zhang2024advancing APACrefauthors Zhang, Y. , Lu, H. , Liu, X. , Chen, Y. Wu, K. APACrefauthors \ 2024 . Advancing generalizable remote physiological measurement through the integration of explicit and implicit prior knowledge Advancing generalizable remote physiological measu...

    2024 arXiv

  81. [89]

    , Chen, C

    zhao2023delta APACrefauthors Zhao, B. , Chen, C. Xia, S T. APACrefauthors \ . DELTA: DEGRADATION-FREE FULLY TEST-TIME ADAPTATION Delta: Degradation-free fully test-time adaptation

  82. [90]

    , Yang, Y

    zhou2024mixstyle APACrefauthors Zhou, K. , Yang, Y. , Qiao, Y. Xiang, T. APACrefauthors \ 2024 . Mixstyle neural networks for domain generalization and adaptation Mixstyle neural networks for domain generalization and adaptation . International Journal of Computer Vision 132 3...

    2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.