Pith. sign in

REVIEW 7 minor 46 references

Chart-based conformal scores for gaze and head pose systematically undercover near singularities even when overall coverage looks correct.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 10:36 UTC pith:RHLXAJT3

load-bearing objection Chart-based conformal scores silently undercover near singularities by 30–50 pp; the impossibility result and controlled experiment make the claim solid, and geodesic scoring is a free fix.

arxiv 2607.02565 v1 pith:RHLXAJT3 submitted 2026-06-29 cs.CV cs.HC

Coordinate Singularities Break Conformal Coverage for Gaze and Head Pose

Mohammadreza Jamalifard , Yaxiong Lei , Parastoo Azizinezhad , Javier Andreu-Perez This is my paper
classification cs.CV cs.HC
keywords conformal predictiongaze estimationhead pose estimationcoordinate singularitiesRiemannian geometrygeodesic scoringSO(3)S2
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Conformal prediction is supposed to give distribution-free reliability guarantees for vision systems, but those guarantees depend on how error is measured. For gaze on the sphere and head pose in SO(3), many pipelines still score residuals in yaw–pitch or Euler charts. Those charts compress distances near poles and gimbal lock, so a single global threshold carves out vanishingly small intrinsic regions there. The paper shows that slice-conditional coverage at a nominal 90 percent target collapses by 30–50 percentage points in those regions across four standard datasets, while marginal coverage stays near target. The failure is structural: scalar adaptive methods can only resize the set, not fix its distorted shape. Switching the nonconformity score to a coordinate-free geodesic (or a monotone equivalent) removes the geometric distortion without retraining and at negligible cost. The practical message is that reliability for manifold-valued outputs is not only a calibration problem; it is also a geometry-of-the-score problem.

Core claim

When the conformal nonconformity score is the Euclidean norm of chart residuals (yaw–pitch L2 or Euler L2), the resulting prediction sets inherit the chart’s metric distortion. Near coordinate singularities the same fixed threshold therefore covers far less manifold volume than it does in well-conditioned regions, redistributing coverage so that poles and near-gimbal-lock poses systematically undercover even though marginal coverage remains correctly calibrated. Geodesic scoring restores intrinsic isotropy and recovers the missing coverage.

What carries the argument

Proposition 2: any acceptance set defined by scalar thresholding of a chart-coordinate residual norm is the pre-image of a chart-space ball; its local axis ratios are fixed by the eigenvalues of the metric tensor and therefore cannot be corrected by any scalar radius adaptation (normalised CP, CQR, etc.). The Riemannian volume density supplies a simple diagnostic that tracks where the collapse occurs.

Load-bearing premise

The first-order linearisation of the chart map and the local tangent-space error model remain qualitatively predictive at the finite radii and extreme angles actually present in the real datasets.

What would settle it

On any of the four datasets, replace the chart-norm score with geodesic scoring while keeping the identical model and calibration split; if near-pole or near-gimbal-lock slice coverage does not rise substantially toward the nominal 90 percent target while marginal coverage stays controlled, the geometric claim is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 7 minor

Summary. The paper shows that conformal prediction for manifold-valued vision outputs (gaze on S², head pose on SO(3)) can suffer large slice-conditional undercoverage near coordinate singularities when nonconformity scores are defined in chart coordinates (yaw–pitch or Euler L2), even when marginal coverage is correctly calibrated near 90%. Across ETH-XGaze, Gaze360, BIWI, and AFLW2000-3D, coverage drops by 30–50 pp near poles and gimbal lock. The authors prove (Proposition 2) that scalar adaptive methods (normalised CP, CQR-style threshold modulation) only rescale chart-space balls and cannot change axis ratios fixed by the metric-tensor eigenvalues. They propose the Riemannian volume density as a diagnostic and show that coordinate-free geodesic (or monotone-equivalent) scoring removes the chart-induced distortion without retraining and with negligible cost. A controlled SO(3) experiment with isotropic Lie-algebra noise isolates score geometry, and a three-layer decomposition separates chart distortion from heteroscedastic model error.

Significance. If the claims hold—and the evidence is strong—this is a practically important and theoretically clean contribution for reliability in geometric vision. Conformal prediction is increasingly used for distribution-free guarantees; the paper identifies a failure mode that is invisible under marginal metrics yet concentrated in safety-relevant regimes (extreme pitch, profile poses). Proposition 2 is an elementary but load-bearing geometric identity that cleanly rules out a natural class of remedies. The controlled BIWI experiment isolates score geometry from model error; the four-dataset audit, backbone ablations, and multi-chart/Mahalanobis comparisons make the empirical case robust. The proposed fix is immediately actionable (no retraining, ≤0.02 µs/sample). Credit is due for the clean impossibility result, the geometry-isolating controlled experiment, the three-layer decomposition, and the practical scoring protocol.

minor comments (7)
  1. In Sec. 4.3 the text reports normalised Euler coverage of 62.8±4.0% on AFLW2000-3D, but this number does not appear in Table 5. Adding it (or a short note) would make the Layer-2 vs Layer-3 comparison fully self-contained in the main table.
  2. Sec. 3.2 (“Validity of the linearisation”) already notes that Prop. 1’s quantitative bounds can be loose at extreme angles. A single clarifying sentence earlier in Sec. 3.2 stating that Prop. 1 supplies local intuition and first-order predictions, while the structural claim rests on Prop. 2 and the finite-radius experiments, would help readers weight the two results correctly.
  3. Fig. 3 is very effective; the shaded “degraded / severe / collapsed” bands are useful. Consider stating the exact coverage thresholds used for those bands in the caption so the figure is fully self-contained.
  4. Sec. 5 mentions GazeTR-ViT near-pole numbers (30.3% YP, 81.3% norm. geodesic) that support the “stronger models amplify distortion” claim. A one-row summary in the main text (or a small table) would avoid forcing readers into the supplement for a headline ablation.
  5. Notation: ρ(ξ) is introduced as volume density and later used as a correlation diagnostic. A brief reminder that for the standard charts ρ reduces to |cos(·)| (already stated) could be repeated once near Tables 2–5 where the ρ–coverage correlations are reported.
  6. Minor typography: “T able” appears with a space in several table captions (e.g., “T able 1”, “T able 2”); fix to “Table”. Also “F unctions” in the Sec. 3.4 heading.
  7. Sec. 6’s practical protocol is clear. A short decision note on when to prefer plain geodesic vs normalised geodesic vs Mondrian (once the base score is intrinsic) would help practitioners operationalise the three-layer decomposition.

Circularity Check

0 steps flagged

No significant circularity: geometric identities and held-out empirical measurements stand independently of any fitted inputs or self-referential definitions.

full rationale

The paper's load-bearing claims do not reduce to their inputs by construction. Proposition 2 is an elementary geometric identity: any acceptance set that is a sublevel set of a monotone function of the chart residual norm is the preimage of a chart ball, whose pullback under the chart differential is an ellipsoid whose axis ratios equal sqrt(lambda_max/lambda_min) of the metric tensor and are therefore independent of the scalar threshold or normalisation function h. This identity does not depend on data, fitted parameters, or linearisation. Proposition 1 supplies only local first-order bounds under a tangent-space error model; the authors themselves flag that the quantitative bounds loosen when the metric varies rapidly, yet the qualitative directional collapse is confirmed by finite-radius experiments. The controlled BIWI experiment injects isotropic Lie-algebra noise by construction, so any coverage variation is attributable solely to score geometry; the real-model tables (ETH-XGaze, Gaze360, AFLW2000-3D) measure slice-conditional coverage on held-out subject-disjoint splits after ordinary split conformal calibration. The volume-density diagnostic is a closed-form geometric quantity (rho = |cos theta| or |cos beta|) whose correlation with coverage is an observed statistic, not a fitted prediction. No uniqueness theorem is imported from the authors' prior work, no ansatz is smuggled via self-citation, and no parameter fitted on one slice is re-presented as a prediction of the same slice. The three-layer decomposition cleanly separates marginal validity, heteroscedastic scale, and chart-induced shape; the persistent gap between normalised chart scores and normalised geodesic scores is therefore an independent empirical confirmation of the geometric claim rather than a circular restatement. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The central claim rests on standard Riemannian geometry, the exchangeability assumption of split conformal prediction, and a local Gaussian tangent-space error model. No free parameters are fitted to produce the coverage numbers; the only modelling choices are the conventional charts (yaw–pitch, ZYX Euler) already used by the community and the definition of the high-pitch / near-gimbal slices. No new physical entities are postulated.

free parameters (2)
  • slice thresholds (|pitch|>70°, |β|>60°)
    Chosen by the authors to mark “near-singularity” regimes; physiologically motivated but still a free design choice that defines the reported collapse magnitudes.
  • k-NN bandwidth for normalised scores
    Difficulty estimate σ̂(x) uses a k-NN mean absolute residual; k is a free hyper-parameter whose exact value is deferred to the supplement.
axioms (4)
  • domain assumption Exchangeability of calibration and test points (standard split conformal assumption)
    Required for the marginal coverage guarantee; stated in Sec. 3 and Limitations.
  • domain assumption Local error model Y = exp_p(ε) with ε ~ N(0, Σ_p) small enough to remain inside the chart domain
    Used to derive the coverage bounds of Proposition 1; acknowledged to be first-order.
  • standard math Metric tensors of the standard yaw–pitch chart on S² and ZYX Euler chart on SO(3)
    Classical differential-geometry facts; G = diag(cos²θ,1) and the 3×3 Euler metric with det G = cos²β.
  • ad hoc to paper Acceptance sets of the form h(∥φ(y)−φ(ŷ)∥₂, x) ≤ q with h non-decreasing
    Defines the class of scalar chart-based conformal predictors to which Proposition 2 applies; excludes learned anisotropic scores.

pith-pipeline@v1.1.0-grok45 · 19591 in / 2939 out tokens · 24823 ms · 2026-07-12T10:36:20.948345+00:00 · methodology

0 comments
read the original abstract

Conformal prediction provides distribution-free reliability guarantees for vision systems, but these guarantees depend on how prediction errors are measured in the output space. Many vision tasks produce outputs on curved spaces (e.g. gaze directions on the sphere or 3D head rotations), yet intermediate prediction heads, residuals, uncertainty estimates, or conformal scores are often defined in flat coordinate charts such as yaw-pitch or Euler angles. We show that this scoring choice introduces systematic geometric distortion near coordinate singularities (large pitch angles on the sphere and poses approaching gimbal lock in 3D rotations). Across four datasets (ETH-XGaze, Gaze360, BIWI, AFLW2000-3D), slice-conditional coverage at a nominal 90% target drops by 30-50 percentage points in these regions, falling to 38.9% on ETH-XGaze and 42.0% on Gaze360 at gaze pitch above 70 degrees, and to 57.5% on BIWI and 55.2% on AFLW2000-3D at head pose pitch above 60 degrees near gimbal lock, despite marginal coverage remaining near 90%. We prove that this is structural. Scalar thresholding changes the size of chart-coordinate prediction sets but leaves their distorted axis ratios unchanged. To diagnose this hidden failure mode, we show that a simple geometric quantity, the Riemannian volume density, strongly correlates with where coverage collapse occurs. Finally, we show that coordinate-free geodesic scoring removes this distortion. It requires no retraining and adds negligible computational cost.

Figures

Figures reproduced from arXiv: 2607.02565 by Javier Andreu-Perez, Mohammadreza Jamalifard, Parastoo Azizinezhad, Yaxiong Lei.

Figure 1
Figure 1. Figure 1: Chart distortion breaks conformal coverage [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Protocol overview. The key decision point is the nonconformity score. Chart￾based scores inherit coordinate distortion; geodesic scores are coordinate-free. The vol￾ume density ρ indicates where chart scores are likely to fail. coordinate residuals. (Section 3.4 defines the full score taxonomy; here we focus on schart and the geodesic score sgeo = dg(y, yˆ).) The discrepancy is captured by the Riemannian m… view at source ↗
Figure 3
Figure 3. Figure 3: Conditional coverage from equator to pole [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

46 extracted references · 4 linked inside Pith

  1. [1]

    In: 2023 8th International Conference on Frontiers of Signal Processing (ICFSP)

    Abdelrahman, A.A., Hempel, T., Khalifa, A., Al-Hamadi, A., Dinges, L.: L2cs- net: Fine-grained gaze estimation in unconstrained environments. In: 2023 8th International Conference on Frontiers of Signal Processing (ICFSP). pp. 98–102. IEEE (2023)

    2023

  2. [2]

    In: Pro- ceedings of the IEEE/CVF Winter Conference on applications of computer vision

    Cantarini, G., Tomenotti, F.F., Noceti, N., Odone, F.: Hhp-net: A light het- eroscedastic neural network for head pose estimation with uncertainty. In: Pro- ceedings of the IEEE/CVF Winter Conference on applications of computer vision. pp. 3521–3530 (2022)

    2022

  3. [3]

    In: 2022 26th International Conference on Pattern Recognition (ICPR)

    Cheng, Y., Lu, F.: Gaze estimation using transformer. In: 2022 26th International Conference on Pattern Recognition (ICPR). pp. 3341–3347. IEEE (2022)

    2022

  4. [4]

    Electronic Journal of Statistics19(2), 6141–6166 (2025)

    Cholaquidis, A., Gamboa, F., Moreno, L.: Conformal inference for regression on riemannian manifolds. Electronic Journal of Statistics19(2), 6141–6166 (2025)

    2025

  5. [5]

    Do Carmo, M.P., Flaherty Francis, J.: Riemannian geometry, vol. 393. Springer (1992)

    1992

  6. [6]

    International journal of computer vision101(3), 437–458 (2013)

    Fanelli, G., Dantone, M., Gall, J., Fossati, A., Van Gool, L.: Random forests for real time 3d face analysis. International journal of computer vision101(3), 437–458 (2013)

    2013

  7. [7]

    Clinical Biomechanics14(7), 462–470 (1999)

    Feipel, V., Rondelet, B., Le Pallec, J.P., Rooze, M.: Normal global motion of the cervical spine:: an electrogoniometric study. Clinical Biomechanics14(7), 462–470 (1999)

    1999

  8. [8]

    Bernoulli29(1), 1–23 (2023)

    Fontana, M., Zeni, G., Vantini, S.: Conformal prediction: a unified review of theory and new challenges. Bernoulli29(1), 1–23 (2023)

    2023

  9. [9]

    Information and Inference: A Journal of the IMA10(2), 455–482 (2021)

    Foygel Barber, R., Candes, E.J., Ramdas, A., Tibshirani, R.J.: The limits of distribution-free conditional predictive inference. Information and Inference: A Journal of the IMA10(2), 455–482 (2021)

    2021

  10. [10]

    IET Computer Vision10(4), 308– 314 (2016)

    Fridman, L., Lee, J., Reimer, B., Victor, T.: ‘owl’and ‘lizard’: Patterns of head pose and eye pose in driver gaze classification. IET Computer Vision10(4), 308– 314 (2016)

    2016

  11. [11]

    In: international conference on machine learn- ing

    Gal, Y., Ghahramani, Z.: Dropout as a bayesian approximation: Representing model uncertainty in deep learning. In: international conference on machine learn- ing. pp. 1050–1059. PMLR (2016)

    2016

  12. [12]

    In: Advances in Neural Information Processing Systems (NeurIPS) (2021)

    Gibbs, I., Candès, E.: Adaptive conformal inference under distribution shift. In: Advances in Neural Information Processing Systems (NeurIPS) (2021)

    2021

  13. [13]

    Cambridge university press (2003)

    Hartley, R., Zisserman, A.: Multiple view geometry in computer vision. Cambridge university press (2003)

    2003

  14. [14]

    In: 2022 IEEE International Conference on image processing (ICIP)

    Hempel, T., Abdelrahman, A.A., Al-Hamadi, A.: 6d rotation representation for unconstrained head pose estimation. In: 2022 IEEE International Conference on image processing (ICIP). pp. 2496–2500. IEEE (2022)

    2022

  15. [15]

    Journal of Math- ematical Imaging and Vision35(2), 155–164 (2009)

    Huynh, D.Q.: Metrics for 3d rotations: Comparison and analysis. Journal of Math- ematical Imaging and Vision35(2), 155–164 (2009)

    2009

  16. [16]

    In: Conformal and Probabilistic Prediction and Applications

    Johnstone, C., Cox, B.: Conformal uncertainty sets for robust optimization. In: Conformal and Probabilistic Prediction and Applications. pp. 72–90. PMLR (2021)

    2021

  17. [17]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Kellnhofer, P., Recasens, A., Stent, S., Matusik, W., Torralba, A.: Gaze360: Physi- cally unconstrained gaze estimation in the wild. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 6912–6921 (2019)

    2019

  18. [18]

    (eds.) Advances Coordinate Singularities Break Conformal Coverage 17 in Neural Information Processing Systems

    Kendall, A., Gal, Y.: What uncertainties do we need in bayesian deep learning for computer vision? In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (eds.) Advances Coordinate Singularities Break Conformal Coverage 17 in Neural Information Processing Systems. vol. 30. Curran Associates, Inc. (2017),https : / ...

    2017

  19. [19]

    In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R

    Lakshminarayanan, B., Pritzel, A., Blundell, C.: Simple and scalable predic- tive uncertainty estimation using deep ensembles. In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (eds.) Ad- vances in Neural Information Processing Systems. vol. 30. Curran Associates, Inc. (2017),https://proceedings.neurips.cc/pa...

    2017

  20. [20]

    Lee, J.M.: Introduction to Riemannian manifolds, vol. 2. Springer (2018)

    2018

  21. [21]

    Eye33(7), 1192 (2019)

    Lee, W.J., Kim, J.H., Shin, Y.U., Hwang, S., Lim, H.W.: Correction: Differences in eye movement range based on age and gaze direction. Eye33(7), 1192 (2019)

    2019

  22. [22]

    Journal of the American Statistical Association 113(523), 1094–1111 (2018)

    Lei, J., G’Sell, M., Rinaldo, A., Tibshirani, R.J., Wasserman, L.: Distribution-free predictive inference for regression. Journal of the American Statistical Association 113(523), 1094–1111 (2018)

    2018

  23. [23]

    ACM Computing Surveys 56(2), 1–38 (2023)

    Lei, Y., He, S., Khamis, M., Ye, J.: An end-to-end review of gaze estimation and its interactive applications on handheld mobile devices. ACM Computing Surveys 56(2), 1–38 (2023)

    2023

  24. [24]

    arXiv preprint arXiv:2502.10570 (2025)

    Lei, Y., Wang, Y., Buchanan, F., Zhao, M., Sugano, Y., He, S., Khamis, M., Ye, J.: Quantifying the impact of motion on 2d gaze estimation in real-world mobile interactions. arXiv preprint arXiv:2502.10570 (2025)

    Pith/arXiv arXiv 2025

  25. [25]

    arXiv preprint arXiv:2505.22769 (2025)

    Lei, Y., Zhao, M., Wang, Y., He, S., Sugano, Y., Khamis, M., Ye, J.: Mac- gaze: Motion-aware continual calibration for mobile gaze tracking. arXiv preprint arXiv:2505.22769 (2025)

    Pith/arXiv arXiv 2025

  26. [26]

    Ma, Y., Soatto, S., Kosecka, J., Sastry, S.S.: An invitation to 3-d vision: from images to geometric models, vol. 26. Springer Science & Business Media (2012)

    2012

  27. [27]

    In: Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition

    Martyniuk, T., Kupyn, O., Kurlyak, Y., Krashenyi, I., Matas, J., Sharmanska, V.: Dad-3dheads: A large-scale dense, accurate and diverse dataset for 3d head alignment from a single image. In: Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition. pp. 20942–20952 (2022)

    2022

  28. [28]

    In: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., Lin, H

    Mohlin, D., Sullivan, J., Bianchi, G.: Probabilistic orientation estimation with ma- trix fisher distributions. In: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., Lin, H. (eds.) Advances in Neural Information Processing Systems. vol. 33, pp. 4884–4893. Curran Associates, Inc. (2020),https://proceedings.neurips.cc/ paper _ files / paper / 2020 / fil...

    2020

  29. [29]

    In: NeuRIPS 2022 Workshop on Gaze Meets ML (2022)

    Nikan, S., Upadhyay, D.: Appearance-based gaze estimation for driver monitoring. In: NeuRIPS 2022 Workshop on Gaze Meets ML (2022)

    2022

  30. [30]

    In: European conference on machine learning

    Papadopoulos, H., Proedrou, K., Vovk, V., Gammerman, A.: Inductive confidence machines for regression. In: European conference on machine learning. pp. 345–356. Springer (2002)

    2002

  31. [31]

    ACM Transactions On Graphics (TOG)35(6), 1–12 (2016)

    Patney, A., Salvi, M., Kim, J., Kaplanyan, A., Wyman, C., Benty, N., Luebke, D., Lefohn, A.: Towards foveated rendering for gaze-tracked virtual reality. ACM Transactions On Graphics (TOG)35(6), 1–12 (2016)

    2016

  32. [32]

    In: Proceedings of the European conference on computer vision (ECCV)

    Prokudin, S., Gehler, P., Nowozin, S.: Deep directional statistics: Pose estimation with uncertainty quantification. In: Proceedings of the European conference on computer vision (ECCV). pp. 534–551 (2018)

    2018

  33. [33]

    In: Ad- vances in Neural Information Processing Systems (NeurIPS) (2019) 18 M

    Romano, Y., Patterson, E., Candès, E.: Conformalized quantile regression. In: Ad- vances in Neural Information Processing Systems (NeurIPS) (2019) 18 M. Jamalifard et al

    2019

  34. [34]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops

    Ruiz, N., Chong, E., Rehg, J.M.: Fine-grained head pose estimation without key- points. In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops. pp. 2074–2083 (2018)

    2074

  35. [35]

    arXiv preprint arXiv: 2107.07511 (2021)

    Stephen, B., et al.: A gentle introduction to conformal prediction and distribution- free uncertainty quantification. arXiv preprint arXiv: 2107.07511 (2021)

    Pith/arXiv arXiv 2021

  36. [36]

    Advances in neural information processing systems32(2019)

    Tibshirani, R.J., Foygel Barber, R., Candes, E., Ramdas, A.: Conformal prediction under covariate shift. Advances in neural information processing systems32(2019)

    2019

  37. [37]

    Springer (2005)

    Vovk, V., Gammerman, A., Shafer, G.: Algorithmic learning in a random world. Springer (2005)

    2005

  38. [38]

    In: Proceedings of the 2025 Symposium on Eye Tracking Research and Applications

    Wang, Y., Yan, R., Lei, Y., Fu, X.: Ptgaze: Cross-domain gaze estimation via proxy tuning. In: Proceedings of the 2025 Symposium on Eye Tracking Research and Applications. pp. 1–2 (2025)

    2025

  39. [39]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Yang, T.Y., Chen, Y.T., Lin, Y.Y., Chuang, Y.Y.: Fsa-net: Learning fine-grained structure aggregation for head pose estimation from a single image. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 1087–1096 (2019)

    2019

  40. [40]

    In: European conference on computer vision

    Zhang, X., Park, S., Beeler, T., Bradley, D., Tang, S., Hilliges, O.: Eth-xgaze: A large scale dataset for gaze estimation under extreme head pose and gaze variation. In: European conference on computer vision. pp. 365–381. Springer (2020)

    2020

  41. [41]

    In: 31st British Machine Vision Conference (BMVC 2020)

    Zhang, X., Sugano, Y., Bulling, A., Hilliges, O.: Learning-based region selection for end-to-end gaze estimation. In: 31st British Machine Vision Conference (BMVC 2020). p. 86. British Machine Vision Association (2020)

    2020

  42. [42]

    arXiv preprint arXiv:2303.10062 (2023)

    Zheng, Q., Zhang, X.: Confidence-aware 3d gaze estimation and evaluation metric. arXiv preprint arXiv:2303.10062 (2023)

    Pith/arXiv arXiv 2023

  43. [43]

    IEEE Transactions on Image Processing33, 2851–2866 (2024)

    Zhong, W., Xia, C., Zhang, D., Han, J.: Uncertainty modeling for gaze estimation. IEEE Transactions on Image Processing33, 2851–2866 (2024)

    2024

  44. [44]

    ACM computing surveys58(2), 1–37 (2025)

    Zhou, X., Chen, B., Gui, Y., Cheng, L.: Conformal prediction: A data perspective. ACM computing surveys58(2), 1–37 (2025)

    2025

  45. [45]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Zhou, Y., Barnes, C., Lu, J., Yang, J., Li, H.: On the continuity of rotation rep- resentations in neural networks. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 5745–5753 (2019)

    2019

  46. [46]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhu, X., Lei, Z., Liu, X., Shi, H., Li, S.Z.: Face alignment across large poses: A 3d solution. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 146–155 (2016)

    2016