Pith. sign in

REVIEW 3 major objections 4 minor 49 references

Relative Age Estimation Using Face Images

T0 review · 3 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read Reference-face comparisons trim age error to 2.47 yr

desk verdict A plausible differential-refinement scheme with consistent small gains, but a test-time retrieval sentence that looks like ground-truth leakage must be clarified before the headline MAE is credible. read the letter →

arxiv 2502.04852 v1 pith:4L73QORZ submitted 2025-02-07 cs.CV

Ran Sandhaus , Yosi Keller This is my paper · ORCID
classification cs.CV
keywords facialageestimationdifferentialregressionreference-basedrefinementsubject-exclusiveprotocolmeanabsoluteerrorMORPHIICACDdeeplearning
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 argues that estimating a person's age from a face photo can be made more accurate by turning the task into a relative one: first a baseline regressor gives a rough age, then a second network estimates how much older or younger the face looks compared to a handful of reference faces of known, similar age and similar appearance, and the weighted average of those differences corrects the baseline. The correction is small in magnitude, so the network learns to predict residuals in a narrow range rather than the full span of human aging. The authors report that this differential, reference-based refinement achieves a mean absolute error of 2.47 years on MORPH II and 5.27 years on CACD under the identity-disjoint subject-exclusive protocol, improving on prior state-of-the-art results. Because the refinement is model-agnostic, any absolute age regressor can in principle be improved by wrapping it in this framework.

What carries the argument

The carrying object is the Differential Age Regressor (DAR), a CNN that consumes the query image and reference images, concatenates their visual embeddings with numeric age embeddings of the known reference ages, and predicts each pairwise age difference as a softmax-weighted mixture of regression heads centered on offsets from $-20$ to $+20$. The final estimate is $\hat{y} = \hat{a}_q + \sum_r w_r d_r$, where $\hat{a}_q$ is the baseline age and $w_r$ are learned per-reference weights. A second-order refinement regresses residuals around each head's center, and training enforces the symmetry constraint $\Delta(x_1,x_2) = -\Delta(x_2,x_1)$ through multi-task losses (MSE, cross-entropy, and mean-variance terms). The retrieval machinery that feeds the DAR uses the baseline prediction for age-based candidate selection, a face-embedding nearest-neighbor step for visual similarity, and kernel density estimation of the baseline's error distribution to sample augmented ages during training.

What would settle it

Re-run the MORPH II subject-exclusive evaluation with reference retrieval driven only by the baseline's predicted age -- never by the ground-truth age -- and compare the resulting MAE to the reported 2.47; a material increase would show the published number depended on test answers.

Watch

Extended reading notes

Core claim

The central claim is that the residual error of a baseline age estimator is easier to model than absolute age itself. The paper's Differential Age Regressor (DAR) takes a query face and a retrieved reference set, predicts the age difference between the query and each reference, learns a weight for each reference, and forms the final estimate as the baseline prediction plus the weighted average of the predicted differences. Retrieval is age-constrained and appearance-based: candidates are faces whose known ages lie near the baseline estimate, then the visually closest are selected via face embeddings, with random sampling for diversity. The paper also models the baseline's out-of-sample error distribution by kernel density estimation and samples from it during training, and shows that iterating this refinement lowers the MORPH II MAE from 2.50 to 2.47. In the authors' experiments the full system outperforms the published subject-exclusive results it compares against on both MORPH II and CACD.

Load-bearing premise

The reported accuracy assumes that at test time the reference faces are selected using the baseline model's predicted age, not the person's true age, and that the error distribution estimated on a small subject-exclusive subset generalizes to the test set.

Editorial extensions

If this is right

  • Any baseline age regressor can be plugged into the differential refinement loop, so the reported gain is not tied to one backbone architecture.
  • Under the subject-exclusive protocol, the paper reports MAE 2.47 on MORPH II and 5.27 on CACD, improving on the listed prior SE-protocol results.
  • The refinement's benefit depends on the quality of the estimated error distribution; KDE sampling outperformed a uniform distribution and converged faster in the ablation (Table VI).
  • Iterating the refinement gives a small but consistent accuracy gain, from 2.50 to 2.47 MAE with the KDE-based configuration (Table VIII).
  • The need for identity labels restricts the method to datasets like MORPH II and CACD where subject-exclusive splits are possible.

Reading between the lines

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

  • Editorial inference: the test-time sentence 'ensuring that $\hat{a}_q = a_q$' is ambiguous. If it means the true age is used to choose reference faces at test time, the reported MAE would not reflect a real deployment; reproducing the protocol requires that only the baseline's predicted age drives retrieval.
  • Editorial inference: the same relative-refinement recipe could be applied to other fine-grained visual attributes where absolute scoring is noisy but pairwise comparisons are reliable, such as apparent health or facial attractiveness.
  • Editorial inference: because the framework samples references from a closed training database, its accuracy at the tails of the age range (60+ on MORPH II) appears to be limited by reference coverage as much as by the regressor, based on Table IX.
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

3 major / 4 minor

Summary. The paper proposes a two-stage framework for facial age estimation. A Baseline Age Regressor (BAR) first produces an initial absolute-age estimate; a Differential Age Regressor (DAR) then refines it by comparing the query face with reference faces of similar age and appearance and estimating age differences. The reference set is retrieved using the BAR estimate and face-embedding nearest-neighbor similarity. Training uses an augmented age aq+epsilon sampled from a KDE-estimated error distribution, and an iterative refinement updates the error distribution over training epochs. Experiments on MORPH II and CACD under the subject-exclusive protocol report MAEs of 2.47 and 5.27, which the authors claim are state-of-the-art, and the paper includes ablations and a bias analysis.

Significance. The central idea of refining an absolute regressor with learned relative-age differences against retrieved reference faces is interesting, timely, and potentially model-agnostic. If the test-time protocol is genuinely deployment-valid, the reported gain over the Hierarchical-Attention BAR is a useful demonstration. The paper has concrete strengths: it uses the subject-exclusive protocol to avoid identity leakage, it isolates components through ablations, it reports results on two public datasets, and it includes a demographic bias analysis. However, the reported gains are small (0.06 MAE over the previous SOTA and 0.03–0.04 in ablations), and the absence of uncertainty estimates means the headline claim is not yet robustly supported. The stress-test concern about the test-time retrieval sentence is valid and must be resolved, because if ground-truth age is used for retrieval the reported MAE is not a valid deployment estimate.

major comments (3)
  1. [Section III-A] The sentence “During testing, the same retrieval process is repeated without using Eq. 1, ensuring that baq = aq” is self-contradictory. Omitting Eq. 1 leaves baq as the BAR’s raw predicted age, not the ground-truth age aq. If the implementation literally sets baq = aq at test time, references are retrieved using the answer, which would invalidate the reported MAE of 2.47. If this is a typo and the intended statement is that the BAR-predicted age is used for retrieval, the main text must say so explicitly; Fig. 2’s caption suggests the latter, but the current sentence is load-bearing and must be corrected.
  2. [Section IV-A / Tables II–VIII] The claimed state-of-the-art result rests on very small MAE differences (2.47 vs 2.53 on MORPH II, and 2.47 vs 2.50 or 2.51 in ablations), yet no error bars, significance tests, or number of repeated runs are reported. Because the reference set is randomly sampled (Section III-A), the method has intrinsic run-to-run variability. Please report mean and standard deviation over multiple seeds and, ideally, paired comparisons against the BAR baseline. Without this, the headline improvement is not statistically established.
  3. [Section III-A / Section IV] Several retrieval details needed for reproducibility and for assessing the method are missing. The size K of the initial age-retrieved pool is never specified, and it is unclear how “their age is baq” is interpreted when baq is continuous: exact integer equality, or an age range, and if so what bound? The KDE bandwidth for estimating Dε is not reported, and the 2% subject-exclusive subset used for the estimate is not analyzed for representativeness. Please specify K, the age tolerance, the KDE bandwidth, and the sensitivity of the final MAE to these choices.
minor comments (4)
  1. [Section III-C, Eq. (5)] The loss La_MSE appears both outside the sum and inside the sum over references, so it is double-counted and the relative weighting in the text is not matched by the equation. Please clarify whether this is a typographical duplication and state the intended loss weighting.
  2. [References] Several author names in the reference list and related-work text appear garbled (e.g., “Wang and Kamikaze” for [10], “Lieu” for [18]/[38], “Rote” for [30], “Tain” for [17]). These should be corrected for professional presentation.
  3. [Section V] The bias analysis would be more useful with a compact numeric table of MAE by gender and ethnicity; currently the text mentions a 24% gender gap and several qualitative claims, but the supporting numbers appear only in figures with no confidence intervals.
  4. [Section IV-B, Tables IV–VIII] The tables contain formatting artifacts such as “2 .47” and “0 .0003”; these should be cleaned up before publication.

Circularity Check

1 steps flagged · score 4.0 of 10

Test-time retrieval sentence in §III-A, read literally, feeds ground-truth age into the refinement; the rest of the derivation is non-circular.

  1. self definitional [Section III-A, 'Reference Images Retrieval', test-phase sentence after Eq. 1]
    "During testing, the same retrieval process is repeated without using Eq. 1, ensuring that baq = aq."

    Eq. 1 defines baq = aq + ε, so omitting Eq. 1 cannot produce baq = aq; it would leave baq as the BAR estimate. Read literally, the test-time retrieval selects references by the ground-truth age, and then Eq. 2 (by = baq + sum_r w_r d_r) builds the final prediction on that target value. The reported MAE of 2.47 would then be a constructed/leaked quantity rather than an independent deployment prediction. The accompanying Fig. 2 and the phrase 'the same retrieval process' suggest the BAR estimate was intended, making this a likely typographical error; but as written it is a load-bearing circular step.

full rationale

Aside from the test-phase sentence, the derivation chain is not circular. The DAR is a learned residual regressor: it is trained with Eq. 5 to predict age differences from image pairs, and the final estimate in Eq. 2 is the BAR estimate plus a learned weighted correction, not a fitted constant. The augmentation distribution Dε is estimated by KDE on a 2% subject-exclusive training subset and is used only for training-time reference sampling, so it does not leak test labels. The comparison includes published external baselines (Coral, Mean-Variance, soft-ranking, DCDL) in addition to the authors' prior work [22], so the self-citation supplies a component and a baseline but is not the sole evidence for the gain. If the ambiguous sentence is corrected to use the BAR prediction, the appropriate circularity score would be 1-2; the literal reading of that sentence is what prevents a clean non-circular verdict.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim rests on standard supervised-learning assumptions plus several face-specific domain assumptions: that age labels are accurate, that VGG16 embedding similarity tracks age-relevant appearance, that a 2% subject-exclusive subset models the BAR error distribution, and that small age differences are easier to regress than absolute ages. No new physical or conceptual entities are introduced; the DAR is a network architecture, not an invented entity in the sense of a new force, particle, or conserved quantity.

free parameters (7)
  • R reference count = 10 (R=2 also gives 2.47)
    Chosen after ablation; the number of reference images directly enters Eqs. 2 and 5 and the retrieval scheme.
  • P nearest-neighbor pool size = 30
    Chosen by hand; controls the size of the visually similar pool from which R references are sampled.
  • C second-order differential range = 20
    The regression heads span -C to C in Eq. 4; chosen by hand and not justified by data.
  • KDE error truncation bound = [-20, +20]
    Section III-A restricts D_epsilon to this interval because KDE estimation was inconsistent outside it; affects Eq. 1 sampling.
  • Iterative refinement steps = 2
    Eq. 6 is applied twice based on the ablation in Table VIII; more iterations were not tested.
  • Distribution estimation subset size = 2% of training set
    The BAR error distribution D_epsilon is fit on this subject-exclusive subset; chosen by hand.
  • KDE bandwidth = not reported
    Needed to reproduce D_epsilon exactly; the paper does not state the bandwidth or selection procedure.
assumptions (6)
  • domain assumption Small age differences between similar-looking faces are easier to regress than absolute age
    Central premise of differential refinement; introduced in Section I and used throughout.
  • domain assumption The 2% subject-exclusive subset yields a representative BAR error distribution D_epsilon
    Section III-A fits KDE on this subset; if it is not representative, reference sampling during training is biased.
  • domain assumption MORPH II and CACD age labels are accurate enough as ground truth
    All MAE numbers and D_epsilon estimates depend on label quality; Section IV explicitly treats these datasets as reliable.
  • domain assumption Age difference symmetry Delta(x1,x2) = -Delta(x2,x1) holds for faces
    Enforced in Section III-C to regularize training; aging appearance is not strictly symmetric.
  • domain assumption VGG16 embedding L2 distance selects references that are visually and age-relevant similar
    Retrieval in Section III-A uses these embeddings; if they capture identity or pose instead of age appearance, refinement signal degrades.
  • standard math Standard supervised learning assumptions: subject-exclusive split, i.i.d. samples, and ImageNet pretraining transfer
    Invoked by the training protocol in Section IV; not stated as axioms but required for the MAE to generalize.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Relative Age Estimation Using Face Images." pith.science (2026). https://pith.science/paper/4L73QORZ

@misc{pith2026250204852,
  author       = {Pith},
  title        = {Pith review of: Relative Age Estimation Using Face Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4L73QORZ}},
  note         = {Machine review of arXiv:2502.04852}
}
read the original abstract

This work introduces a novel deep-learning approach for estimating age from a single facial image by refining an initial age estimate. The refinement leverages a reference face database of individuals with similar ages and appearances. We employ a network that estimates age differences between an input image and reference images with known ages, thus refining the initial estimate. Our method explicitly models age-dependent facial variations using differential regression, yielding improved accuracy compared to conventional absolute age estimation. Additionally, we introduce an age augmentation scheme that iteratively refines initial age estimates by modeling their error distribution during training. This iterative approach further enhances the initial estimates. Our approach surpasses existing methods, achieving state-of-the-art accuracy on the MORPH II and CACD datasets. Furthermore, we examine the biases inherent in contemporary state-of-the-art age estimation techniques.

Figures

Figures reproduced from arXiv: 2502.04852 by the authors.

Figure 1
Figure 1. Differential age estimation. The Baseline Age Re￾gressor (BAR) estimates the age baq of the input image xq. xbq is the CNN embedding of xq used with baq to retrieve the set of reference images that are of age baq and most visually similar to xq. The ages of the reference images are known. The Differential Age Regressor (DAR) estimates the age differences between xq and the reference images and uses them to refine ba… view at source ↗
Figure 2
Figure 2. Differential age estimation in the training and test phases. In training time, the groundtruth age aq of the input face image xq, is augmented using Eq. 1 and used to retrieve the reference set of images {xr} R 1 . In the test phase, a Baseline Age Regressor (BAR) estimates aˆq, the age of xq that is used to retrieve {xr} R 1 . III. DIFFERENTIAL AGE ESTIMATION We propose a novel framework for facial age estimation, … view at source ↗
Figure 3
Figure 3. The Differential Age Regressor (DAR) network. The embeddings of the query image xq and the reference images {xr} R 1 are computed by a CNN, while the initial age estimate of the query image aˆq and reference ages {ar} R 1 are encoded by an embedding layer. The embeddings are concatenated to xˆq and {xˆr} R 1 . The DAR network uses the embeddings to estimate the age differences {dr} R 1 and the weights per reference … view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Reference faces retrieval. (a) The age aq of xq is used to retrieve {xbr} K 1 the embeddings of the reference images of the same age. (b) xbq, the embedding of xq is used to retrieve {xbr} P 1 , P ≪ K, the P embeddings in {xbr} K 1 closest to xbq. (c) We randomly sampl…
Figure 5
Figure 5. Figure 5: Differential Age Estimation. This network estimates the age differential between the query image xˆq and a single reference image xˆ. Their embeddings are concatenated as pairs  xˆq, xˆ i r [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Distribution of mean error for the proposed method, [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Distribution of mean absolute error for the proposed [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8 [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 11
Figure 11. Figure 11: Gender bias (MAE histogram). A gender-wise distribution of MAE for our proposed method, over the Morph II dataset. White Black Hispanic Asian Ethnicity 1.0 0.8 0.6 0.4 0.2 0.0 Mean Estimation Error [age] gender Male Female [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 12
Figure 12. Figure 12: Gender and ethnicity bias. A per-ethnicity and gender breakdown of the mean error for our proposed method, using the Morph II dataset. 0 2 4 6 8 10 12 14 Mean Absolute Estimation Error [age] 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Probability ethnicity White Black Hispanic As…
Figure 13
Figure 13. Figure 13: Gender bias (MAE histogram). A per-ethnicity and gender breakdown of MAE for our proposed method, using the Morph II dataset [PITH_FULL_IMAGE:figures/full_fig_p009_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 47 canonical work pages

  1. [1]

    Video analytics for business intelligence,

    A. Hakeem, H. Gupta, A. Kanaujia, T. E. Choe, K. Gunda, A. W. Scanlon, L. Yu, Z. Zhang, P. L. Venetianer, Z. Rasheed, and N. Haering, “Video analytics for business intelligence,” in Video Analytics for Business Intelligence, 2012

    work page 2012

  2. [2]

    Comparing different classifiers for automatic age estimation,

    A. Lanitis, C. Draganova, and C. Christodoulou, “Comparing different classifiers for automatic age estimation,” IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), vol. 34, pp. 621–628, 2004

    work page 2004

  3. [3]

    Facial aging: A 54-year, three-dimensional population study,

    V . Lambros, “Facial aging: A 54-year, three-dimensional population study,” Plastic and reconstructive surgery , vol. 145, pp. 921–928, 04 2020

    work page 2020

  4. [4]

    Age and gender estimation of unfiltered faces,

    E. Eidinger, R. Enbar, and T. Hassner, “Age and gender estimation of unfiltered faces,” IEEE Transactions on Information Forensics and Security, vol. 9, no. 12, pp. 2170–2179, 2014

    work page 2014

  5. [5]

    Simultaneous dimensionality reduction and human age estimation via kernel partial least squares regression,

    G. Guo and G. Mu, “Simultaneous dimensionality reduction and human age estimation via kernel partial least squares regression,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR), June 2011, pp. 657–664

    work page 2011

  6. [6]

    Ordinal hyperplanes ranker with cost sensitivities for age estimation,

    K.-Y . Chang, C.-S. Chen, and Y .-P. Hung, “Ordinal hyperplanes ranker with cost sensitivities for age estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2011, pp. 585–592

    work page 2011

  7. [7]

    Human age estimation using ranking svm,

    D. Cao, Z. Lei, Z. Zhang, J. Feng, and S. Z. Li, “Human age estimation using ranking svm,” in Biometric Recognition , W.-S. Zheng, Z. Sun, Y . Wang, X. Chen, P. C. Yuen, and J. Lai, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2012, pp. 324–331

    work page 2012

  8. [8]

    Gender classification in large databases,

    E. Ram ´on-Balmaseda, J. Lorenzo-Navarro, and M. Castrill ´on-Santana, “Gender classification in large databases,” in Progress in Pattern Recog- nition, Image Analysis, Computer Vision, and Applications . Springer, 2012, pp. 74–81

    work page 2012

Show all 49 references
  1. [9]

    Human age estimation: What is the influence across race and gender?

    G. Guo and G. Mu, “Human age estimation: What is the influence across race and gender?” in IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), June 2010, pp. 71–78

    2010

  2. [10]

    Age estimation via unsupervised neural networks,

    X. Wang and C. Kambhamettu, “Age estimation via unsupervised neural networks,” in International Conference on Automatic Face and Gesture Recognition (FGR), vol. 1, May 2015, pp. 1–6

    2015

  3. [11]

    Cumulative attribute space for age and crowd density estimation,

    K. Chen, S. Gong, T. Xiang, and C. Loy, “Cumulative attribute space for age and crowd density estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2013, pp. 2467–2474

    2013

  4. [12]

    Age and gender classification using convolu- tional neural networks,

    G. Levi and T. Hassner, “Age and gender classification using convolu- tional neural networks,” in IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), June 2015, pp. 34–42

    2015

  5. [13]

    Label-sensitive deep metric learning for facial age estimation,

    H. Liu, J. Lu, J. Feng, and J. Zhou, “Label-sensitive deep metric learning for facial age estimation,” IEEE Transactions on Information Forensics and Security, vol. 13, no. 2, pp. 292–305, 2018

    2018

  6. [14]

    Deepage: Deep learning of face-based age estimation,

    O. Sendik and Y . Keller, “Deepage: Deep learning of face-based age estimation,” Signal Processing: Image Communication, vol. 78, 08 2019

    2019

  7. [15]

    Dimensionality reduction by learning an invariant mapping,

    R. Hadsell, S. Chopra, and Y . LeCun, “Dimensionality reduction by learning an invariant mapping,” in Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR) , vol. 2, 2006, pp. 1735–1742

    2006

  8. [16]

    Deep cost- sensitive and order-preserving feature learning for cross-population age estimation,

    K. Li, J. Xing, C. Su, W. Hu, Y . Zhang, and S. Maybank, “Deep cost- sensitive and order-preserving feature learning for cross-population age estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2018, pp. 399–408

    2018

  9. [17]

    Relationships self-learning based gender-aware age estimation,

    Q. Tian, M. Cao, S. Chen, and H. Yin, “Relationships self-learning based gender-aware age estimation,” Neural Processing Letters, vol. 50, no. 3, pp. 2141–2160, 2019

    2019

  10. [18]

    Using ranking-cnn for age estimation,

    S. Chen, C. Zhang, M. Dong, J. Le, and M. Rao, “Using ranking-cnn for age estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 742–751

    2017

  11. [19]

    Ordinal regression with multiple output cnn for age estimation,

    Z. Niu, M. Zhou, L. Wang, X. Gao, and G. Hua, “Ordinal regression with multiple output cnn for age estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 4920–4928

    2016

  12. [20]

    Soft-ranking label encoding for robust facial age estimation,

    X. Zeng, J. Huang, and C. Ding, “Soft-ranking label encoding for robust facial age estimation,” IEEE Access, vol. 8, pp. 134 209–134 218, 2020

    2020

  13. [21]

    Distilling ordinal relation and dark knowledge for facial age estimation,

    Q. Zhao, J. Dong, H. Yu, and S. Chen, “Distilling ordinal relation and dark knowledge for facial age estimation,” IEEE Transactions on Neural Networks and Learning Systems , vol. PP, 07 2020

    2020

  14. [22]

    Hierarchical Attention-Based Age Estimation and Bias Analysis ,

    S. Hiba and Y . Keller, “ Hierarchical Attention-Based Age Estimation and Bias Analysis ,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 12, pp. 14 682–14 692, Dec. 2023

    2023

  15. [23]

    Improving face-based age esti- mation with attention-based dynamic patch fusion,

    H. Wang, V . Sanchez, and C.-T. Li, “Improving face-based age esti- mation with attention-based dynamic patch fusion,” IEEE Trans. Image Process., vol. 31, pp. 1084–1096, 2022

    2022

  16. [24]

    Fp-age: Leveraging face parsing attention for facial age estimation in the wild,

    Y . Lin, J. Shen, Y . Wang, and M. Pantic, “Fp-age: Leveraging face parsing attention for facial age estimation in the wild,” 2021

    2021

  17. [25]

    P-age: Pexels dataset for robust spatio-temporal apparent age classification,

    A. Ali, A. Marisetty, and F. Br ´emond, “P-age: Pexels dataset for robust spatio-temporal apparent age classification,” in IEEE Workshop on Applications of Computer Vision , January 2024, pp. 8606–8615

    2024

  18. [26]

    Morph: a longitudinal image database of normal adult age-progression,

    K. Ricanek and T. Tesafaye, “Morph: a longitudinal image database of normal adult age-progression,” inInternational Conference on Automatic Face and Gesture Recognition (FGR) , 2006, pp. 341–345

    2006

  19. [27]

    Cross-age reference coding for age-invariant face recognition and retrieval,

    B.-C. Chen, C.-S. Chen, and W. H. Hsu, “Cross-age reference coding for age-invariant face recognition and retrieval,” in Proceedings of the European Conference on Computer Vision (ECCV) , D. Fleet, T. Pajdla, B. Schiele, and T. Tuytelaars, Eds. Cham: Springer International Pub...

    2014

  20. [28]

    Age estimation from face images: Human vs. machine performance,

    H. Han, C. Otto, and A. Jain, “Age estimation from face images: Human vs. machine performance,” in International Conference on Biometrics , June 2013, pp. 1–8

    2013

  21. [29]

    Gender shades: Intersectional accuracy disparities in commercial gender classification,

    J. Buolamwini and T. Gebru, “Gender shades: Intersectional accuracy disparities in commercial gender classification,” in Proceedings of the 1st Conference on Fairness, Accountability and Transparency , S. A. Friedler and C. Wilson, Eds., vol. 81, 2018, pp. 77–91

    2018

  22. [30]

    Dex: Deep expectation of apparent age from a single image,

    R. Rothe, R. Timofte, and L. Van Gool, “Dex: Deep expectation of apparent age from a single image,” in Proceedings of the IEEE International Conference on Computer Vision Workshops (ICCVW) , 2015, pp. 252–257

    2015

  23. [31]

    Mean-variance loss for deep age estimation from a face,

    H. Pan, H. Han, S. Shan, and X. Chen, “Mean-variance loss for deep age estimation from a face,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE Computer Society, 2018, pp. 5285–5294

    2018

  24. [32]

    Deep label distribution learning for apparent age estimation,

    X. Yang, B. Gao, C. Xing, Z. Huo, X. Wei, Y . Zhou, J. Wu, and X. Geng, “Deep label distribution learning for apparent age estimation,” in Proceedings of the IEEE International Conference on Computer Vision Workshops (ICCVW), 2015, pp. 344–350

    2015

  25. [33]

    Deep regression forests for age estimation,

    W. Shen, Y . Guo, Y . Wang, K. Zhao, B. Wang, and A. Yuille, “Deep regression forests for age estimation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 2304–2313

    2018

  26. [34]

    Bridgenet: A continuity-aware probabilistic network for age estimation,

    W. Li, J. Lu, J. Feng, C. Xu, J. Zhou, and Q. Tian, “Bridgenet: A continuity-aware probabilistic network for age estimation,” in Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 1145–1154

    2019

  27. [35]

    Rank consistent ordinal regres- sion for neural networks with application to age estimation,

    W. Cao, V . Mirjalili, and S. Raschka, “Rank consistent ordinal regres- sion for neural networks with application to age estimation,” Pattern Recognition Letters, vol. 140, pp. 325–331, 2020

    2020

  28. [36]

    Deep conditional distribution learning for age estimation,

    H. Sun, H. Pan, H. Han, and S. Shan, “Deep conditional distribution learning for age estimation,” IEEE Transactions on Information Foren- sics and Security , vol. 16, pp. 4679–4690, 2021

    2021

  29. [37]

    General vs. long- tailed age estimation: An approach to kill two birds with one stone,

    Z. Bao, Z. Tan, J. Li, J. Wan, X. Ma, and Z. Lei, “General vs. long- tailed age estimation: An approach to kill two birds with one stone,” IEEE Trans. Image Process. , vol. 32, p. 6155–6167, Jan. 2023

    2023

  30. [38]

    Siamese graph learning for semi-supervised age estimation,

    H. Liu, M. Ma, Z. Gao, Z. Deng, F. Li, and Z. Li, “Siamese graph learning for semi-supervised age estimation,” IEEE Trans. Multimedia , vol. 25, pp. 9586–9596, 2023

    2023

  31. [39]

    Daa: A delta age adain operation for age estimation via binary code transformer,

    P. Chen, X. Zhang, Y . Li, J. Tao, B. Xiao, B. Wang, and Z. Jiang, “Daa: A delta age adain operation for age estimation via binary code transformer,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2023

    2023

  32. [40]

    The FERET database and evaluation procedure for face-recognition algorithms,

    P. Phillips, H. Wechsler, J. Huang, and P. J. Rauss, “The FERET database and evaluation procedure for face-recognition algorithms,” Image and Vision Computing, vol. 16, no. 5, pp. 295–306, 1998

    1998

  33. [41]

    The fg-net aging database,

    T. Cootes and A. Lanitis, “The fg-net aging database,” 2002, available online at http://www-prima.inrialpes.fr/FGnet/

    2002

  34. [42]

    Apparent and real age estimation in still images with deep residual regressors on appa-real database,,

    E. Agustsson, R. Timofte, S. Escalera, X. Baro, I. Guyon, and R. Rothe., “Apparent and real age estimation in still images with deep residual regressors on appa-real database,,” in International Conference on Au- tomatic Face and Gesture Recognition (FGR) . IEEE, 2017

    2017

  35. [43]

    Age progres- sion/regression by conditional adversarial autoencoder,

    Zhang, Zhifei, Song, Yang, Qi, and Hairong, “Age progres- sion/regression by conditional adversarial autoencoder,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR). IEEE, 2017, pp. 4352–4360

    2017

  36. [44]

    Morph: a longitudinal image database of normal adult age-progression,

    K. Ricanek and T. Tesafaye, “Morph: a longitudinal image database of normal adult age-progression,” inInternational Conference on Automatic Face and Gesture Recognition (FGR) , April 2006, pp. 341–345

    2006

  37. [45]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in Proceedings of the International Con- ference on Learning Representations (ICLR) , Y . Bengio and Y . LeCun, Eds., 2015. 11

    2015

  38. [46]

    B. W. Silverman, Density Estimation for Statistics and Data Analysis , ser. Monographs on Statistics and Applied Probability. London: Chapman and Hall, 1986

    1986

  39. [47]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, J. Yang, N. Xue, I. Kotsia, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” IEEE Transac- tions on Pattern Analysis and Machine Intelligence , vol. 44, no. 10, p. 5962–5979, Oct. 2022

    2022

  40. [48]

    Ranger21: a synergistic deep learning optimizer,

    L. Wright and N. Demeure, “Ranger21: a synergistic deep learning optimizer,” CoRR, vol. abs/2106.13731, 2021. [Online]. Available: https://arxiv.org/abs/2106.13731

    2021 arXiv

  41. [49]

    Face recognition: too bias, or not too bias?

    J. P. Robinson, G. Livitz, Y . Henon, C. Qin, Y . Fu, and S. Timoner, “Face recognition: too bias, or not too bias?” in IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) , 2020, pp. 0–10

    2020

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.