Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

MAC-Gaze: Motion-Aware Continual Calibration for Mobile Gaze Tracking

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A phone's own motion sensors can decide when gaze tracking needs recalibration, and replay-based continual learning makes the fix stick, cutting average error from 2.81 cm to 1.92 cm on naturalistic data.

desk verdict The core idea is solid and worth refereeing, but the main accuracy claim is internally inconsistent: Table 3 reports 1.41 cm while Table 7, the replay-ratio sweep on the same dataset, never goes below 1.59 cm. read the letter →

arxiv 2505.22769 v3 pith:QJE7KP2N submitted 2025-05-28 cs.HC cs.CV

Yaxiong Lei , Mingyue Zhao , Yuheng Wang , Shijing He , Yusuke Sugano , Mohamed Khamis , Juan Ye This is my paper
classification cs.HCcs.CV
keywords gazeestimationmobiletrackingcontinualcalibrationIMUactivityrecognitionreplay-basedlearninghybridtriggerMotioncatastrophicforgetting
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

Mobile gaze tracking on smartphones drifts as users shift posture, change how they hold the device, or move while walking; one-off calibration cannot keep up. The paper aims to show that calibration can be made continual and automatic: a phone's IMU sensors detect when the user's motion state changes, and a replay-based continual learner updates the gaze calibrator without forgetting earlier motion conditions. The reported result is that this MAC-Gaze system lowers average gaze error from 1.73 cm to 1.41 cm on the RGBDGaze dataset and from 2.81 cm to 1.92 cm on the newly collected MotionGaze dataset, compared with traditional one-off calibration. A sympathetic reader would care because this is a practical route to keeping gaze interfaces usable during everyday phone interaction without repeated manual calibration.

What carries the argument

The load-bearing mechanism is the calibration trigger plus the replay buffer working as a pair. The hybrid trigger first applies temporal majority voting to the supervised activity predictions of the IMU model, so a single misclassification does not fire a recalibration; it then scores each raw IMU reading by its maximum likelihood under a Gaussian mixture model built from previously stored motion samples, and declares a new fine-grained motion pattern when the outlier ratio in the voting window passes a threshold. When a trigger fires, the continual calibration module updates a two-layer MLP using a loss that supervises both the new calibration samples and replayed samples from the buffer, which prevents catastrophic forgetting. The paper reports that removing the buffer raises error from 1.41 cm to 1.73 cm, and replacing the hybrid trigger with the classifier alone raises it to 1.58 cm, which is the direct evidence that both halves of the mechanism carry the result.

What would settle it

A recording with continuous gaze ground truth and IMU data in which a user makes large gaze shifts while holding the phone perfectly still would settle the trigger's necessity: if gaze error rises sharply with no IMU deviation, the motion signal is missing the relevant change. The converse test, IMU deviations that produce no rise in gaze error, would reveal over-triggering.

Watch

Extended reading notes

Core claim

The paper's central claim is that the two open questions in mobile gaze calibration—when to recalibrate and how to recalibrate—can be answered together with motion sensing and replay-based continual learning. For 'when,' MAC-Gaze combines supervised activity recognition (an IMU encoder-decoder with a classification head) with unsupervised Gaussian mixture clustering on raw IMU samples; recalibration is triggered when the stable classified activity changes or when the current IMU stream largely falls outside previously seen GMM components. For 'how,' a small MLP calibrator is retrained on new calibration data together with a memory buffer of samples from earlier motion states, balancing adaptation with retention. On RGBDGaze and MotionGaze, the full system outperforms no calibration, one-off calibration, and classifier-only triggering, and on RGBDGaze it also outperforms an oracle that knows every motion transition in advance.

Load-bearing premise

The load-bearing premise is that phone motion sensed by the IMU is a reliable indicator of when the head-eye-camera relationship behind gaze error has changed, a link the paper adopts from prior work rather than validating directly per participant.

Editorial extensions

If this is right

  • Motion-triggered recalibration reduces average 2D gaze error from 1.73 cm to 1.41 cm on RGBDGaze and from 2.81 cm to 1.92 cm on MotionGaze, beating one-off calibration in every tested posture.
  • When recalibration happens matters as much as how often: giving participants the same number of calibrations at fixed time intervals yields 1.71 cm versus 1.41 cm for motion-triggered calibration.
  • Replay-based learning is what prevents forgetting: without the memory buffer, error jumps from 1.41 cm to 1.73 cm.
  • The calibration approach wraps around different gaze backbones; with iTracker as the base model, error drops by 22.0% on RGBDGaze and 40.1% on MotionGaze.
  • Starting posture does not change the outcome: one-way ANOVA across four initial calibration postures gives p = 0.782, meaning the continual learner adapts regardless of task order.

Reading between the lines

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

  • A direct test of the paper's implicit causal claim would compare gaze error and IMU outlier scores continuously; if error rises before the IMU trigger fires, the motion signal is lagging the true domain shift.
  • The choice to cluster raw IMU readings rather than learned latent features implies that the activity-classification representation discards exactly the micro-motion information calibration needs, which motivates future work on hierarchical or multi-scale representations that preserve both.
  • An extension the paper leaves implicit is combining the motion trigger with implicit visual signals such as touch, reading, or on-screen content so that users who change gaze behavior without changing phone motion can still trigger recalibration.
  • The per-participant fluctuation seen for highly dynamic users suggests that motion-predictive models, rather than reactive triggers, are the next step for the most active device-handling patterns.
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 / 5 minor

Summary. The paper presents MAC-Gaze, a motion-aware continual calibration framework for mobile gaze tracking. It combines a pre-trained visual gaze estimator with an IMU-based activity recognition model and a hybrid trigger that uses supervised classification plus GMM-based outlier detection to decide when to recalibrate. Calibration updates are performed with replay-based continual learning to avoid catastrophic forgetting. The system is evaluated on the RGBDGaze dataset and a newly collected MotionGaze dataset, reporting 19.9% and 31.7% error reductions over traditional calibration baselines, together with ablations and a generalization study using iTracker as the backbone.

Significance. If the quantitative claims hold, the work is significant: it addresses a practically important gap in mobile gaze tracking—deciding when to trigger recalibration and how to update the calibrator without forgetting previous motion conditions—and it contributes a new multimodal dataset with IMU, image, and gaze data. The ablations isolating the hybrid trigger, replay, and motion-based timing are useful, as is the demonstration that the approach generalizes to a second backbone. The main limitation is that the headline number is not internally reproducible from the reported tables, which must be resolved before the claimed improvement can be assessed.

major comments (4)
  1. [Section 5.4, Table 7; Section 5.1, Table 3; Abstract] Section 5.4 reports a replay-ratio sweep for MAC-Gaze on RGBDGaze whose best average error is 1.59 cm at a 70% replay ratio, yet Table 3 and Table 6 report 1.41 cm for the full MAC-Gaze system, and no cell in Table 7 reaches 1.41 cm. The paper does not state which replay ratio or protocol difference produces the 1.41 cm result. This makes the headline 19.9% improvement internally unreproducible; the abstract's 1.73 cm baseline also matches neither the 1.74 cm one-off calibration row nor the 1.76 cm no-calibration row in Table 3. If the correct value were 1.59 cm, the improvement over one-off calibration would be only about 8.6%, materially changing the central claim.
  2. [Section 5.1, Oracle Motion-Aware definition] The Oracle Motion-Aware baseline trains a separate calibration model for each motion using only 10% of that motion's data, with no replay or accumulation across motions. It is therefore not an upper bound in terms of available training data: MAC-Gaze uses a growing replay buffer containing data from all previously encountered motions. Beating the oracle (1.41 vs 1.54 cm on RGBDGaze; 1.92 vs 2.23 cm on MotionGaze) may reflect the oracle's smaller calibration set rather than superior triggering. The oracle should be given matched replay/accumulation, or MAC-Gaze should be compared with equal total calibration data, to support the claim that MAC-Gaze outperforms perfect motion detection.
  3. [Section 4.4.3 and Section 5.4] The grid search over detection thresholds is described as 'select the best setting' without stating whether the selection uses the held-out validation subset or the test participants; Section 3.4 says thresholds were determined on a held-out validation subset, but Section 4.4.3 does not confirm this or report the selected values. Similarly, Section 5.4 sweeps replay ratios on RGBDGaze and identifies 70% as best, but does not say whether the main results in Table 3 use that ratio or a different one. Please report the selected threshold and replay-ratio values, specify the validation split, and clarify the relationship between Table 7 and Table 3; otherwise the reported error reductions are open to outcome-dependent selection.
  4. [Section 5.2, Section 5.3, Figure 7] The 'when to calibrate' contribution asserts that IMU-detected motion changes are a reliable proxy for gaze-error degradation, but the paper only reports trigger counts and qualitative timelines. There is no quantitative evaluation of whether triggered recalibrations coincide with measured increases in gaze error, or of false-trigger rates relative to error degradation; the matched-count time-based comparison in Table 6 is a useful start but does not provide precision/recall of the trigger. Figure 7 shows participants p11 and p17 whose error increases after recalibration, which the paper attributes to temporal mismatch, but no aggregate statistic is given. Please add a per-trigger or per-segment analysis to validate the association between detected motion changes and actual calibration need.
minor comments (5)
  1. [Section 3.4 and Section 4.4.3] Section 3.4 uses τ for the consensus ratio and τ_i for the GMM likelihood threshold, but Section 4.4.3 refers to 'the two thresholds for outlier detection τ and ρ' and never defines ρ; the threshold notation and the selected values should be aligned.
  2. [Figure 7 and Eq. (2)] Figure 7 labels the y-axis 'MSE(cm)' and the text calls the metric MSE, while Eq. (2) and the rest of the paper use mean Euclidean error; please unify the terminology.
  3. [Section 6.3 and Keywords] There are a few typos: 'more fine-course' should likely be 'more fine-grained', and the keyword list contains 'Continue Learning' instead of 'Continual Learning'.
  4. [Figure 5 caption] The caption of Figure 5 refers to 'True Label' and 'Weak Label' while the text describes ground-truth versus classifier predictions; the distinction between 'Weak Label' and classifier prediction is not explained.
  5. [Table 5] Table 5 reports average data length per participant (3,869 frames for RGBDGaze, 17,521 for MotionGaze) without stating units or whether frames are evenly distributed across postures, which would aid interpretation of the calibration-frequency comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No definitional circularity: MAC-Gaze reports held-out empirical errors; replay-ratio and table inconsistencies are correctness issues, not circular reductions.

full rationale

MAC-Gaze's central claims are empirical measurements on held-out test participants, not predictions derived from fitted parameters. The calibrator MC is trained on a 10% calibration subset of each detected motion and evaluated on the remaining 90% (Section 4.3), so the reported 1.41 cm and 1.92 cm errors are genuine out-of-sample results. Trigger thresholds are stated to be tuned on a held-out validation subset of the training data (Section 3.4), which is a standard model-selection procedure and not a definitional loop. The only self-citation used as a data source, MotionGaze [39], is not load-bearing: the dataset is described in detail in Section 4.1.2 and is paired with the independent public RGBDGaze benchmark. The discrepancy between Table 3/6 (1.41 cm) and Table 7 (best 1.59 cm at 70% replay) is a serious internal-consistency and reproducibility problem, and selecting the replay ratio on the test set would be data snooping, but neither is a circular reduction: Table 7's best value is not identical to the headline value, so the claimed improvement does not equal its own input by construction. The assumed causal link from IMU changes to gaze error is a validity assumption, not a circular step.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The method depends on several empirical choices and domain assumptions rather than a formal derivation. The free parameters are mostly tuned thresholds and architecture choices; the key domain assumption is that IMU signals predict gaze-relevant domain shifts. No new physical entities are introduced.

free parameters (7)
  • alpha (replay loss weight) = not reported
    In Eq. 1, alpha balances L_calibration and L_replay; no value or tuning procedure is given in the paper.
  • thresholds (tau, tau_i, outlier ratio) = not reported (grid searched in [0.75, 0.95])
    Section 4.4.3 says grid search selects the best setting on a held-out validation subset; the chosen values are not disclosed.
  • replay ratio = 70%
    Section 5.4 reports that 70% yields lowest error (1.59 cm) and appears to be selected on the test set; this number is used in the main system description, yet the main result reports 1.41 cm, an inconsistency.
  • r (reconstruction/classification balance) = 0.3
    Fixed in Section 3.2 loss weighting without sensitivity analysis.
  • GMM component count K = 10
    Figure 3b shows 10 clusters identified; how K was selected is not described.
  • IMU window size and overlap = 200 timestamps, 75% overlap
    Selected empirically, no ablation on window size.
  • majority voting window size c = 4 s
    Section 5.3 tests 2, 4, 6-8 s and chooses 4 s; no performance table for this choice.
assumptions (5)
  • domain assumption IMU signals are a reliable indicator of the gaze-relevant domain shift (head-eye-camera geometry changes)
    The whole trigger mechanism depends on IMU patterns predicting gaze accuracy degradation; cited prior work [39] supports this, but it is assumed without formal justification in this paper.
  • domain assumption Replay-based continual learning prevents catastrophic forgetting in this setting
    Standard continual learning assumption; supported by ablation but no comparison to other CL methods.
  • domain assumption The 10% of data used for calibration is representative of the motion condition
    Used in the evaluation protocol for initialization and recalibration; if the 10% window is unrepresentative, results would vary.
  • domain assumption The evaluation sequence order does not affect final accuracy
    Section 5.4 reports ANOVA showing no significant difference between starting conditions, so the paper assumes task order invariance.
  • domain assumption Raw IMU vectors, rather than learned features, form separable clusters corresponding to calibration-relevant motion patterns
    Section 3.4 and Figure 3b justify this empirically on one participant (p11); whether it holds across all users is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MAC-Gaze: Motion-Aware Continual Calibration for Mobile Gaze Tracking." pith.science (2026). https://pith.science/paper/QJE7KP2N

@misc{pith2026250522769,
  author       = {Pith},
  title        = {Pith review of: MAC-Gaze: Motion-Aware Continual Calibration for Mobile Gaze Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QJE7KP2N}},
  note         = {Machine review of arXiv:2505.22769}
}
read the original abstract

Mobile gaze tracking faces a fundamental challenge: maintaining accuracy as users naturally change their postures and device orientations. Traditional calibration approaches, like one-off, fail to adapt to these dynamic conditions, leading to degraded performance over time. We present MAC-Gaze, a Motion-Aware continual Calibration approach that leverages smartphone Inertial measurement unit (IMU) sensors and continual learning techniques to automatically detect changes in user motion states and update the gaze tracking model accordingly. Our system integrates a pre-trained visual gaze estimator and an IMU-based activity recognition model with a clustering-based hybrid decision-making mechanism that triggers recalibration when motion patterns deviate significantly from previously encountered states. To enable accumulative learning of new motion conditions while mitigating catastrophic forgetting, we employ replay-based continual learning, allowing the model to maintain performance across previously encountered motion conditions. We evaluate our system through extensive experiments on the publicly available RGBDGaze dataset and our own 10-hour multimodal MotionGaze dataset (481K+ images, 800K+ IMU readings), encompassing a wide range of postures under various motion conditions including sitting, standing, lying, and walking. Results demonstrate that our method reduces gaze estimation error by 19.9% on RGBDGaze (from 1.73 cm to 1.41 cm) and by 31.7% on MotionGaze (from 2.81 cm to 1.92 cm) compared to traditional calibration approaches. Our framework provides a robust solution for maintaining gaze estimation accuracy in mobile scenarios.

Figures

Figures reproduced from arXiv: 2505.22769 by the authors.

Figure 1
Figure 1. Head-eye-camera configurations in handheld systems that consist of up to 15-DoF from head, eye, and device [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Workflow of Motion-Aware Continual Learning Calibration for Mobile Gaze Tracking [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Comparison of feature representations for GMM clustering of fine-grained motion patterns, visualized via [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Comparison of acceleration ranges between RGBDGaze and MotionGaze in box plots [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Temporal visualization of calibration trigger events on the RGBDGaze dataset for a participant. The upper [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: Temporal visualization of calibration trigger events for a participant in the MotionGaze dataset (Lying First [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: Temporal visualization of gaze estimation performance across test segments for multiple participants on the [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Coordinate Singularities Break Conformal Coverage for Gaze and Head Pose

    cs.CV 2026-06 accept novelty 7.0 of 10

    Yaw–pitch and Euler conformal scores redistribute coverage near coordinate singularities; geodesic scores restore slice-conditional reliability without retraining.

Reference graph

Works this paper leans on

72 extracted references · 66 canonical work pages · cited by 1 Pith paper

  1. [1]

    A supervised autoencoder for human activity recognition with inertial sensors

    AN, J., KWON, Y.,ANDCHO, Y.-S. A supervised autoencoder for human activity recognition with inertial sensors. In2023 IEEE International Conference on Big Data (BigData)(2023), IEEE, pp. 1–5

    work page 2023

  2. [2]

    Rgbdgaze: Gaze tracking on smartphones with rgb and depth data

    ARAKAWA, R., GOEL, M., HARRISON, C.,ANDAHUJA, K. Rgbdgaze: Gaze tracking on smartphones with rgb and depth data. InProceedings of the 2022 International Conference on Multimodal Interaction(2022), pp. 329–336. 20 Motion-Aware Continual Calibration for Mobile Gaze TrackingA PREPRINT

    work page 2022

  3. [3]

    Learning gaze-aware compositional gan from limited annotations.Proceedings of the ACM on Computer Graphics and Interactive Techniques 7, 2 (2024), 1–17

    ARANJUELO, N., HUANG, S., ARGANDA-CARRERAS, I., UNZUETA, L., OTAEGUI, O., PFISTER, H.,AND WEI, D. Learning gaze-aware compositional gan from limited annotations.Proceedings of the ACM on Computer Graphics and Interactive Techniques 7, 2 (2024), 1–17

    work page 2024

  4. [4]

    P.,ANDDEWANGAN, D

    BANJAREY, K., SAHU, S. P.,ANDDEWANGAN, D. K. A survey on human activity recognition using sensors and deep learning methods. In2021 5th international conference on computing methodologies and communication (ICCMC)(2021), IEEE, pp. 1610–1617

    work page 2021

  5. [5]

    Adaptive feature fusion network for gaze tracking in mobile tablets

    BAO, Y., CHENG, Y., LIU, Y.,ANDLU, F. Adaptive feature fusion network for gaze tracking in mobile tablets. In2020 25th International Conference on Pattern Recognition (ICPR)(2021), IEEE, pp. 9936–9943

    work page 2021

  6. [6]

    Heterogeneity activity recognition

    BLUNCK, H., BHATTACHARYA, S., PRENTOW, T., KJRGAARD, M.,ANDDEY, A. Heterogeneity activity recognition. UCI Machine Learning Repository, 2015. DOI: https://doi.org/10.24432/C5689X

    doi:10.24432/c5689x 2015

  7. [7]

    Class-incremental continual learning into the extended der-verse.IEEE transactions on pattern analysis and machine intelligence 45, 5 (2022), 5497–5512

    BOSCHINI, M., BONICELLI, L., BUZZEGA, P., PORRELLO, A.,ANDCALDERARA, S. Class-incremental continual learning into the extended der-verse.IEEE transactions on pattern analysis and machine intelligence 45, 5 (2022), 5497–5512

    work page 2022

  8. [8]

    A tutorial on human activity recognition using body-worn inertial sensors.ACM Computing Surveys (CSUR) 46, 3 (2014), 1–33

    BULLING, A., BLANKE, U.,ANDSCHIELE, B. A tutorial on human activity recognition using body-worn inertial sensors.ACM Computing Surveys (CSUR) 46, 3 (2014), 1–33

    work page 2014

Show all 72 references
  1. [9]

    Gazeswipe: Enhancing mobile touchscreen reachability through seamless gaze and finger-swipe integration.arXiv preprint arXiv:2503.21094(2025)

    CAI, Z., HONG, J., WANG, Z.,ANDLU, F. Gazeswipe: Enhancing mobile touchscreen reachability through seamless gaze and finger-swipe integration.arXiv preprint arXiv:2503.21094(2025)

    2025 arXiv

  2. [10]

    CAO, L., WANG, Y., ZHANG, B., JIN, Q.,ANDVASILAKOS, A. V. Gchar: An efficient group-based con- text—aware human activity recognition on smartphone.Journal of Parallel and Distributed Computing 118 (2018), 67–80

    2018

  3. [11]

    Offset calibration for appearance-based gaze estimation via gaze decomposition

    CHEN, Z.,ANDSHI, B. Offset calibration for appearance-based gaze estimation via gaze decomposition. In Proceedings of the IEEE/CVF winter conference on applications of computer vision(2020), pp. 270–279

    2020

  4. [12]

    CHEN, Z.,ANDSHI, B. E. Towards high performance low complexity calibration in appearance based gaze estimation.IEEE transactions on pattern analysis and machine intelligence 45, 1 (2022), 1174–1188

    2022

  5. [13]

    Puregaze: Purifying gaze feature for generalizable gaze estimation

    CHENG, Y., BAO, Y.,ANDLU, F. Puregaze: Purifying gaze feature for generalizable gaze estimation. In Proceedings of the AAAI Conference on Artificial Intelligence(2022), vol. 36, pp. 436–443

    2022

  6. [14]

    Gaze estimation using transformer

    CHENG, Y.,ANDLU, F. Gaze estimation using transformer. In2022 26th International Conference on Pattern Recognition (ICPR)(2022), IEEE, pp. 3341–3347

    2022

  7. [15]

    Dvgaze: Dual-view gaze estimation

    CHENG, Y.,ANDLU, F. Dvgaze: Dual-view gaze estimation. InProceedings of the IEEE/CVF International Conference on Computer Vision(2023), pp. 20632–20641

    2023

  8. [16]

    Appearance-based gaze estimation with deep learning: A review and benchmark.IEEE Transactions on Pattern Analysis and Machine Intelligence(2024)

    CHENG, Y., WANG, H., BAO, Y.,ANDLU, F. Appearance-based gaze estimation with deep learning: A review and benchmark.IEEE Transactions on Pattern Analysis and Machine Intelligence(2024)

    2024

  9. [17]

    Head-mounted eye gaze tracking devices: An overview of modern devices and recent advances.Journal of rehabilitation and assistive technologies engineering 5(2018), 2055668318773991

    COGNOLATO, M., ATZORI, M.,ANDMÜLLER, H. Head-mounted eye gaze tracking devices: An overview of modern devices and recent advances.Journal of rehabilitation and assistive technologies engineering 5(2018), 2055668318773991

    2018

  10. [18]

    A continual learning survey: Defying forgetting in classification tasks.IEEE transactions on pattern analysis and machine intelligence 44, 7 (2021), 3366–3385

    DELANGE, M., ALJUNDI, R., MASANA, M., PARISOT, S., JIA, X., LEONARDIS, A., SLABAUGH, G.,AND TUYTELAARS, T. A continual learning survey: Defying forgetting in classification tasks.IEEE transactions on pattern analysis and machine intelligence 44, 7 (2021), 3366–3385

    2021

  11. [19]

    Combining 3d human pose estimation and imu sensors for human identification and tracking in multi-person environments.IEEE Sensors Letters(2024)

    DEMARCHI, M., TURETTA, C., PRAVADELLI, G.,ANDBOMBIERI, N. Combining 3d human pose estimation and imu sensors for human identification and tracking in multi-person environments.IEEE Sensors Letters(2024)

    2024

  12. [20]

    T.,ANDDUCHOWSKI, A

    DUCHOWSKI, A. T.,ANDDUCHOWSKI, A. T.Eye tracking methodology: Theory and practice. Springer, 2017

    2017

  13. [21]

    Lightweight transformers for human activity recognition on mobile devices.arXiv preprint arXiv:2209.11750(2022)

    EK, S., PORTET, F.,ANDLALANDA, P. Lightweight transformers for human activity recognition on mobile devices.arXiv preprint arXiv:2209.11750(2022)

    2022 arXiv

  14. [22]

    A survey on deep learning for human activity recognition.ACM Computing Surveys (CSUR) 54, 8 (2021), 1–34

    GU, F., CHUNG, M.-H., CHIGNELL, M., VALAEE, S., ZHOU, B.,ANDLIU, X. A survey on deep learning for human activity recognition.ACM Computing Surveys (CSUR) 54, 8 (2021), 1–34

    2021

  15. [23]

    W.,ANDJI, Q

    HANSEN, D. W.,ANDJI, Q. In the eye of the beholder: A survey of models for eyes and gaze.IEEE transactions on pattern analysis and machine intelligence 32, 3 (2009), 478–500

    2009

  16. [24]

    Distilling the knowledge in a neural network.arXiv preprint arXiv:1503.02531(2015)

    HINTON, G. Distilling the knowledge in a neural network.arXiv preprint arXiv:1503.02531(2015)

    2015 arXiv

  17. [25]

    Squeeze-and-excitation networks

    HU, J., SHEN, L.,ANDSUN, G. Squeeze-and-excitation networks. InProceedings of the IEEE conference on computer vision and pattern recognition(2018), pp. 7132–7141. 21 Motion-Aware Continual Calibration for Mobile Gaze TrackingA PREPRINT

    2018

  18. [26]

    Swl-adapt: An unsupervised domain adaptation model with sample weight learning for cross-user wearable human activity recognition

    HU, R., CHEN, L., MIAO, S.,ANDTANG, X. Swl-adapt: An unsupervised domain adaptation model with sample weight learning for cross-user wearable human activity recognition. InProceedings of the AAAI Conference on artificial intelligence(2023), vol. 37, pp. 6012–6020

    2023

  19. [27]

    X., KWOK, T

    HUANG, M. X., KWOK, T. C., NGAI, G., CHAN, S. C.,ANDLEONG, H. V. Building a personalized, auto- calibrating eye tracker from user interactions. InProceedings of the 2016 CHI conference on human factors in computing systems(2016), pp. 5169–5179

    2016

  20. [28]

    Tabletgaze: dataset and analysis for unconstrained appearance-based gaze estimation in mobile tablets.Machine Vision and Applications 28, 5 (2017), 445–461

    HUANG, Q., VEERARAGHAVAN, A.,ANDSABHARWAL, A. Tabletgaze: dataset and analysis for unconstrained appearance-based gaze estimation in mobile tablets.Machine Vision and Applications 28, 5 (2017), 445–461

    2017

  21. [29]

    K.,ANDKO, J

    HUYNH, S., BALAN, R. K.,ANDKO, J. imon: Appearance-based gaze tracking system on mobile devices.Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 5, 4 (dec 2022)

    2022

  22. [30]

    Continual learning in sensor-based human activity recognition: An empirical benchmark analysis.Information Sciences 575(2021), 1–21

    JHA, S., SCHIEMER, M., ZAMBONELLI, F.,ANDYE, J. Continual learning in sensor-based human activity recognition: An empirical benchmark analysis.Information Sciences 575(2021), 1–21

    2021

  23. [31]

    Variational deep embedding: An unsupervised and generative approach to clustering.arXiv preprint arXiv:1611.05148(2016)

    JIANG, Z., ZHENG, Y., TAN, H., TANG, B.,ANDZHOU, H. Variational deep embedding: An unsupervised and generative approach to clustering.arXiv preprint arXiv:1611.05148(2016)

    2016 arXiv

  24. [32]

    Understanding face and eye visibility in front-facing cameras of smartphones used in the wild

    KHAMIS, M., BAIER, A., HENZE, N., ALT, F.,ANDBULLING, A. Understanding face and eye visibility in front-facing cameras of smartphones used in the wild. InCHI ’18(2018), ACM, p. 1–12

    2018

  25. [33]

    KING, D. E. Dlib-ml: A machine learning toolkit.The Journal of Machine Learning Research 10(2009), 1755–1758

    2009

  26. [34]

    Eyemu interactions: Gaze+ imu gestures on mobile devices

    KONG, A., AHUJA, K., GOEL, M.,ANDHARRISON, C. Eyemu interactions: Gaze+ imu gestures on mobile devices. InProceedings of the 2021 International Conference on Multimodal Interaction(2021), pp. 577–585

    2021

  27. [35]

    Eye tracking for everyone

    KRAFKA, K., KHOSLA, A., KELLNHOFER, P., KANNAN, H., BHANDARKAR, S., MATUSIK, W.,AND TORRALBA, A. Eye tracking for everyone. InProceedings of the IEEE conference on computer vision and pattern recognition(2016), pp. 2176–2184

    2016

  28. [36]

    R., WEISS, G

    KWAPISZ, J. R., WEISS, G. M.,ANDMOORE, S. A. Activity recognition using cell phone accelerometers.ACM SigKDD Explorations Newsletter 12, 2 (2011), 74–82

    2011

  29. [37]

    D.,ANDLABRADOR, M

    LARA, O. D.,ANDLABRADOR, M. A. A survey on human activity recognition using wearable sensors.IEEE communications surveys & tutorials 15, 3 (2012), 1192–1209

    2012

  30. [38]

    An end-to-end review of gaze estimation and its interactive applications on handheld mobile devices.ACM Comput

    LEI, Y., HE, S., KHAMIS, M.,ANDYE, J. An end-to-end review of gaze estimation and its interactive applications on handheld mobile devices.ACM Comput. Surv. 56, 2 (sep 2023)

    2023

  31. [39]

    Quantifying the impact of motion on 2d gaze estimation in real-world mobile interactions.arXiv preprint arXiv:2502.10570 (2025)

    LEI, Y., WANG, Y., BUCHANAN, F., ZHAO, M., SUGANO, Y., HE, S., KHAMIS, M.,ANDYE, J. Quantifying the impact of motion on 2d gaze estimation in real-world mobile interactions.arXiv preprint arXiv:2502.10570 (2025)

    2025 arXiv

  32. [40]

    LEI, Y., WANG, Y., CASLIN, T., WISOWATY, A., ZHU, X., KHAMIS, M.,ANDYE, J. Dynamicread: Exploring robust gaze interaction methods for reading on handheld mobile devices under dynamic conditions.Proceedings of the ACM on Human-Computer Interaction 7, ETRA23 (5 2023), 17

    2023

  33. [41]

    Ep n p: An accurate o (n) solution to the p n p problem

    LEPETIT, V., MORENO-NOGUER, F.,ANDFUA, P. Ep n p: An accurate o (n) solution to the p n p problem. International journal of computer vision 81(2009), 155–166

    2009

  34. [42]

    Calibread: Unobtrusive eye tracking calibration from natural reading behavior.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 8, 4 (2024), 1–30

    LIU, C., YU, C., WANG, X., JIANG, J., YANG, T., TANG, B., SHI, Y., LIANG, C.,ANDSHI, Y. Calibread: Unobtrusive eye tracking calibration from natural reading behavior.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 8, 4 (2024), 1–30

    2024

  35. [43]

    3d human pose estimation with single image and inertial measurement unit (imu) sequence.Pattern Recognition 149(2024), 110175

    LIU, L., YANG, J., LIN, Y., ZHANG, P.,ANDZHANG, L. 3d human pose estimation with single image and inertial measurement unit (imu) sequence.Pattern Recognition 149(2024), 110175

    2024

  36. [44]

    J., HU, C.,ANDQIN, X

    LU, W., WANG, J., CHEN, Y., PAN, S. J., HU, C.,ANDQIN, X. Semantic-discriminative mixup for generalizable sensor-based cross-domain activity recognition.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6, 2 (2022), 1–19

    2022

  37. [45]

    Bi-deepvit: Binarized transformer for efficient sensor-based human activity recognition.IEEE Transactions on Mobile Computing(2025)

    LUO, F., LI, A., KHAN, S., WU, K.,ANDWANG, L. Bi-deepvit: Binarized transformer for efficient sensor-based human activity recognition.IEEE Transactions on Mobile Computing(2025)

    2025

  38. [46]

    MA, H., ZHANG, Z., LI, W.,ANDLU, S. Unsupervised human activity representation learning with multi-task deep clustering.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 5, 1 (2021), 1–25

    2021

  39. [47]

    Efficient deep clustering of human activities and how to improve evaluation

    MAHON, L.,ANDLUKASIEWICZ, T. Efficient deep clustering of human activities and how to improve evaluation. InAsian Conference on Machine Learning(2023), PMLR, pp. 722–737. 22 Motion-Aware Continual Calibration for Mobile Gaze TrackingA PREPRINT

    2023

  40. [48]

    Separable self-attention for mobile vision transformers.arXiv preprint arXiv:2206.02680(2022)

    MEHTA, S.,ANDRASTEGARI, M. Separable self-attention for mobile vision transformers.arXiv preprint arXiv:2206.02680(2022)

    2022 arXiv

  41. [49]

    S.,ANDPLÖTZ, T

    MURAHARI, V. S.,ANDPLÖTZ, T. On attention models for human activity recognition. InProceedings of the 2018 ACM international symposium on wearable computers(2018), pp. 100–103

    2018

  42. [50]

    Comparing dwell time, pursuits and gaze gestures for gaze interaction on handheld mobile devices

    NAMNAKANI, O., ABDRABOU, Y., GRIZOU, J., ESTEVES, A.,ANDKHAMIS, M. Comparing dwell time, pursuits and gaze gestures for gaze interaction on handheld mobile devices. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems(2023), pp. 1–17

    2023

  43. [51]

    Stretch gaze targets out: Experimenting with target sizes for gaze-enabled interfaces on mobile devices

    NAMNAKANI, O., ABDRABOU, Y., GRIZOU, J.,ANDKHAMIS, M. Stretch gaze targets out: Experimenting with target sizes for gaze-enabled interfaces on mobile devices. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems(2025), pp. 1–16

    2025

  44. [52]

    D., MOLCHANOV, P., IQBAL, U., HILLIGES, O.,ANDKAUTZ, J

    PARK, S., MELLO, S. D., MOLCHANOV, P., IQBAL, U., HILLIGES, O.,ANDKAUTZ, J. Few-shot adaptive gaze estimation. InProceedings of the IEEE/CVF international conference on computer vision(2019), pp. 9368–9377

    2019

  45. [53]

    A., LUCKYARNO, Y

    SATRIAWAN, A., HERMAWAN, A. A., LUCKYARNO, Y. F.,ANDYUN, J.-H. Predicting future eye gaze using inertial sensors.IEEE Access 11(2023), 67482–67497

    2023

  46. [54]

    P., SHARMA, M

    SINGH, S. P., SHARMA, M. K., LAY-EKUAKILLE, A., GANGWAR, D.,ANDGUPTA, S. Deep convlstm with self-attention for human activity decoding using wearable sensors.IEEE Sensors Journal 21, 6 (2020), 8575–8582

    2020

  47. [55]

    S., KJÆRGAARD, M

    STISEN, A., BLUNCK, H., BHATTACHARYA, S., PRENTOW, T. S., KJÆRGAARD, M. B., DEY, A., SONNE, T., ANDJENSEN, M. M. Smart devices are different: Assessing and mitigatingmobile sensing heterogeneities for activity recognition. InProceedings of the 13th ACM conference on embedded n...

    2015

  48. [56]

    Self-calibrating head-mounted eye trackers using egocentric visual saliency

    SUGANO, Y.,ANDBULLING, A. Self-calibrating head-mounted eye trackers using egocentric visual saliency. In Proceedings of the 28th Annual ACM Symposium on User Interface Software & Technology(2015), pp. 363–372

    2015

  49. [57]

    Learning-by-synthesis for appearance-based 3d gaze estimation

    SUGANO, Y., MATSUSHITA, Y.,ANDSATO, Y. Learning-by-synthesis for appearance-based 3d gaze estimation. InProceedings of the IEEE conference on computer vision and pattern recognition(2014), pp. 1821–1828

    2014

  50. [58]

    An incremental learning method for unconstrained gaze estimation

    SUGANO, Y., MATSUSHITA, Y., SATO, Y.,ANDKOIKE, H. An incremental learning method for unconstrained gaze estimation. InComputer Vision–ECCV 2008: 10th European Conference on Computer Vision, Marseille, France, October 12-18, 2008, Proceedings, Part III 10(2008), Springer, pp. 656–667

    2008

  51. [59]

    Appearance-based gaze estimation with online calibration from mouse operations.IEEE Transactions on Human-Machine Systems 45, 6 (2015), 750–760

    SUGANO, Y., MATSUSHITA, Y., SATO, Y.,ANDKOIKE, H. Appearance-based gaze estimation with online calibration from mouse operations.IEEE Transactions on Human-Machine Systems 45, 6 (2015), 750–760

    2015

  52. [60]

    S.,ANDCHATTOPADHYAY, S

    THAKUR, D., BISWAS, S., HO, E. S.,ANDCHATTOPADHYAY, S. Convae-lstm: Convolutional autoencoder long short-term memory network for smartphone-based human activity recognition.IEEE Access 10(2022), 4137–4156

    2022

  53. [61]

    TONSEN, M., STEIL, J., SUGANO, Y.,ANDBULLING, A. Invisibleeye: Mobile eye tracking using multiple low-resolution cameras and learning-based gaze estimation.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 1, 3 (2017), 1–21

    2017

  54. [62]

    Accelerating eye movement research via accurate and affordable smartphone eye tracking.Nature communications 11, 1 (2020), 4553

    VALLIAPPAN, N., DAI, N., STEINBERG, E., HE, J., ROGERS, K., RAMACHANDRAN, V., XU, P., SHO- JAEIZADEH, M., GUO, L., KOHLHOFF, K.,ET AL. Accelerating eye movement research via accurate and affordable smartphone eye tracking.Nature communications 11, 1 (2020), 4553

    2020

  55. [63]

    A comprehensive survey of continual learning: Theory, method and application.IEEE Transactions on Pattern Analysis & Machine Intelligence, 01 (feb 5555), 1–20

    WANG, L., ZHANG, X., SU, H.,ANDZHU, J. A comprehensive survey of continual learning: Theory, method and application.IEEE Transactions on Pattern Analysis & Machine Intelligence, 01 (feb 5555), 1–20

  56. [64]

    WEISS, G. M. Wisdm smartphone and smartwatch activity and biometrics dataset.UCI Machine Learning Repository: WISDM Smartphone and Smartwatch Activity and Biometrics Dataset Data Set 7, 133190-133202 (2019), 5

    2019

  57. [65]

    Pytorch image models

    WIGHTMAN, R. Pytorch image models. https://github.com/rwightman/pytorch-image-models , 2019

    2019

  58. [66]

    Self-supervised learning for complex activity recognition through motif identification learning

    XIA, Q., MORALES, J., HUANG, Y., HARA, T., WU, K., OSHIMA, H., FUKUDA, M., NAMIOKA, Y.,AND MAEKAWA, T. Self-supervised learning for complex activity recognition through motif identification learning. IEEE Transactions on Mobile Computing(2024)

    2024

  59. [67]

    N., SAN, P

    YANG, J., NGUYEN, M. N., SAN, P. P., LI, X.,ANDKRISHNASWAMY, S. Deep convolutional neural networks on multichannel time series for human activity recognition. InIjcai(2015), vol. 15, Buenos Aires, Argentina, pp. 3995–4001

    2015

  60. [68]

    vgaze: Implicit saliency-aware calibration for continuous gaze tracking on mobile devices

    YANG, S., HE, Y.,ANDJIN, M. vgaze: Implicit saliency-aware calibration for continuous gaze tracking on mobile devices. InIEEE INFOCOM 2021 - IEEE Conference on Computer Communications(2021), pp. 1–10. 23 Motion-Aware Continual Calibration for Mobile Gaze TrackingA PREPRINT

    2021

  61. [69]

    T., YU, B., MENGSHOEL, O

    ZENG, M., NGUYEN, L. T., YU, B., MENGSHOEL, O. J., ZHU, J., WU, P.,ANDZHANG, J. Convolutional neural networks for human activity recognition using mobile sensors. In6th international conference on mobile computing, applications and services(2014), IEEE, pp. 197–205

    2014

  62. [70]

    Revisiting data normalization for appearance-based gaze estimation

    ZHANG, X., SUGANO, Y.,ANDBULLING, A. Revisiting data normalization for appearance-based gaze estimation. InProceedings of the 2018 ACM Symposium on Eye Tracking Research & Applications(New York, NY , USA, 2018), ETRA ’18, Association for Computing Machinery

    2018

  63. [71]

    Appearance-based gaze estimation in the wild

    ZHANG, X., SUGANO, Y., FRITZ, M.,ANDBULLING, A. Appearance-based gaze estimation in the wild. In CVPR ’15(June 2015), IEEE, pp. 4511–4520

    2015

  64. [72]

    Uncertainty modeling for gaze estimation.IEEE Transactions on Image Processing(2024)

    ZHONG, W., XIA, C., ZHANG, D.,ANDHAN, J. Uncertainty modeling for gaze estimation.IEEE Transactions on Image Processing(2024). 24

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.