Pith. sign in

REVIEW 3 major objections 5 minor 67 references

Prior-based Objective Inference Mining Potential Uncertainty for Facial Expression Recognition

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A two-network design that infers objective emotion distributions from action-unit priors can resolve crowd-annotation ambiguity in facial expression recognition, the paper claims, reporting the highest accuracy in its comparisons on five…

desk verdict A solid, well-ablated FER method whose 'objective inference' story is overstated, because the AU pseudolabels are derived from the very emotion labels it claims to correct. read the letter →

arxiv 2411.13024 v1 pith:PJ3FUFJU submitted 2024-11-20 cs.CV

Hanwei Liu , Huiling Cai , Qingcheng Lin , Xuefeng Li , Hui Xiao This is my paper · ORCID
classification cs.CV
keywords FacialexpressionrecognitionAnnotationambiguityactionunitsObjectiveinferenceLabelnoiserobustnessUncertaintyestimationKnowledgedistillation
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

The paper proposes POI, a two-branch network that treats the subjectivity of crowd-sourced expression labels as a missing objective signal and tries to recover that signal from facial action-unit (AU) priors. In the Prior Inference Network, four facial subregion features are supervised by fixed AU-emotion correlation pseudolabels, then pooled through a gated objective inference module into a soft emotion distribution. The Target Recognition Network is trained simultaneously on the original one-hot annotation and on that soft distribution, weighted by an uncertainty score that is high when the subregions disagree. The claim is that this reduces annotation ambiguity: POI reports accuracy of 90.51% on RAF-DB, 63.02% on AffectNet, 90.21% on FERPlus, 99.89% on CK+, and 59.86% on SFEW 2.0, and higher accuracy than SCN, DMUE, LRN, and RUL under 10% to 30% synthetic label noise on three datasets. At inference only the target network runs, so the extra machinery costs nothing at test time.

What carries the argument

The load-bearing mechanism is the pair formed by the Prior Inference Network and the Uncertainty Estimation Module. The PIN splits the face into eye and mouth subregions, flips the right-side crops for symmetry, feeds them into two prior branches, and supervises each branch with AU pseudolabels generated from a fixed AU-expression correlation table, for example happiness associated with AU6 and AU12/AU26 and surprise with AU1/AU2/AU5/AU26. The Objective Inference Module temperature-smooths each subregion's emotion prediction, weights them with a learned gate, and sums them into an intermediate distribution, while a KL divergence term makes every subregion learn from this aggregate, reducing both over-reliance on uncertain priors and over-reliance on noisy labels. The Target Recognition Network's loss mixes cross-entropy on the one-hot annotation with a KL term toward the aggregate soft label, and the UEM sets the mixing weight per image from the variance of subregion predictions. The chain that carries the argument is: muscle-movement prior, then subregion AU features, then a gated soft label, then uncertainty-weighted knowledge transfer into the target network.

What would settle it

Permute the AU-emotion table or replace the pseudolabels with random labels while keeping everything else fixed, and retrain on RAF-DB with 30% flipped labels: if accuracy stays near the reported 85.30%, the prior is not carrying the gain, and if it falls toward the 75.12% baseline, the prior is load-bearing. A complementary check is to compare the PIN's inferred soft distribution with the human label distribution in FERPlus: the objective-inference claim stands or falls on whether the two match.

Watch

Extended reading notes

Core claim

The paper's central claim is that annotation ambiguity in facial expression recognition can be reduced by constructing a relatively objective emotion distribution from facial muscle-movement priors and then training a recognition network on both the subjective one-hot label and that inferred distribution. The Prior Inference Network (PIN) crops four facial subregions, learns action-unit features under pseudolabels derived from a fixed AU-emotion correlation table, and aggregates the subregion predictions through a gated Objective Inference Module into an intermediate soft label. The Target Recognition Network (TRN) is supervised by the original label and by this soft label, with the Uncertainty Estimation Module (UEM) scaling the soft-label term by the disagreement among subregion predictions: high disagreement means low confidence and a larger soft-label weight. The paper reports that this scheme yields the best accuracies in its comparisons on RAF-DB (90.51%), AffectNet (63.02%), FERPlus (90.21%), CK+ (99.89%), and SFEW 2.0 (59.86%), and that it degrades less than the compared noise-robust methods when 10-30% of training labels are flipped.

Load-bearing premise

The load-bearing premise is that the fixed action-unit/emotion correlation table and the smoothed AU pseudolabels derived from possibly noisy emotion labels are objective enough to yield soft labels closer to the true emotion distribution than the original one-hot annotations are; if the prior is wrong for in-the-wild faces, or the noise corrupts the pseudolabels, the soft labels simply inherit the annotation bias.

Editorial extensions

If this is right

  • POI reports 90.51% on RAF-DB, 63.02% on AffectNet, 90.21% on FERPlus, 99.89% on CK+, and 59.86% on SFEW 2.0, the top numbers in the paper's comparisons on those benchmarks.
  • On RAF-DB, FERPlus, and AffectNet with 10-30% flipped labels, POI beats SCN, DMUE, LRN, and RUL; at 30% noise it leads RUL by 3.24, 1.68, and 2.85 points, respectively.
  • The uncertainty score is predictive of accuracy: on RAF-DB the 30% highest-confidence test images reach 99.12% while the 30% lowest-confidence reach 74.32%, a 24.8-point gap, and AffectNet shows a 21.73-point gap.
  • At deployment the PIN is removed, so the test-time cost is just the target network, and no AU annotations or AU detectors are needed at inference.
  • The objective inference module adds 1.24 to 2.67 points over directly using subregion predictions as soft labels under 10-30% synthetic noise.

Reading between the lines

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

  • The same architecture pattern — prior-driven subregion inference yielding soft labels, then uncertainty-weighted transfer to a target network — is portable to other tasks where annotations are subjective and a mechanistic prior exists, such as pathology grading or content moderation.
  • The subregion-consistency confidence could be exposed at test time as a review flag: low-consistency faces are precisely the cases where a single emotion label is least trustworthy, and routing them to human adjudication is a direct deployment extension.
  • Because the soft labels are generated from a fixed action-unit/emotion table, learning or calibrating that mapping per dataset is a natural follow-up experiment that would separate the contribution of the prior itself from the contribution of the gated mutual-learning and uncertainty reweighting scheme.
  • Comparing the inferred soft distributions with the human label distributions available in FERPlus would directly test how objective the inferred labels are; the paper's reported accuracies do not by themselves show this agreement.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes Prior-based Objective Inference (POI), a two-branch network for facial expression recognition (FER) that aims to mitigate annotation ambiguity. The Prior Inference Network (PIN) uses a fixed AU-expression correlation table (Table I) to generate AU pseudolabels from the emotion label (Eq. 4), learns per-subregion features, and produces an intermediate emotion distribution via gated aggregation and mutual learning (Eqs. 5--8). The Target Recognition Network (TRN) is trained with cross-entropy on the original labels and KL divergence toward the PIN's soft labels, weighted by an uncertainty estimate based on subregion prediction consistency (Eqs. 9--11). Only the TRN is used at inference. Experiments on CK+, RAF-DB, AffectNet, FERPlus, and SFEW 2.0, plus synthetic label-noise versions of three datasets, report competitive or state-of-the-art accuracy.

Significance. The paper's empirical scope is broad: comparisons on five public benchmarks, synthetic-noise experiments at 10--30% corruption, component ablations, and visualizations. The reported gains over prior noise-robust methods are consistent under synthetic random flips, and the commitment to release code is a strength. However, the central conceptual claim -- that POI infers 'relatively objective' emotional distributions that correct subjective annotation bias -- is undermined by the fact that the AU pseudolabels are derived deterministically from the very emotion labels the method aims to correct. If reframed as a AU-prior-regularized FER method, the engineering contribution is still interesting, but as written the paper overclaims its objectivity. The experimental methodology also lacks error bars, significance tests, and a clear separation between validation and test sets for hyperparameter selection, which tempers the strength of the benchmark claims.

major comments (3)
  1. [III-C, Eq. (4), Table I; III-D, Eq. (11); Table III] The 'objective' soft labels are not independent of the subjective annotations. In Eq. (4), the AU pseudolabel \hat{y}_{i,n,m} is set to 1-epsilon or epsilon depending only on whether the AU appears in the row of Table I for the image's emotion label y_i. Consequently, the AU supervision, the subregion emotion predictions p_{i,n} trained by Eq. (5), the intermediate prediction \tilde{p}^*_i in Eq. (7), and the soft labels distilled into the TRN in Eq. (11) are all deterministic functions of the same y_i the method claims to correct. If y_i is wrong or systematically biased, the pseudolabel targets are wrong in the same direction. The synthetic-noise experiments only use random flips, which are partially absorbable by the epsilon smoothing and the mutual-learning KL loss, so robustness to random flips does not demonstrate objectivity. The 'w/o inpre' ablation in Table III removes the intermediate prediction but still trains on the same y_i-derived AU targets and the same CE loss on y_i, so it cannot separate the contribution of AU prior knowledge from label-derived regularization. The paper should either use genuinely independent AU supervision (e.g., an external AU detector or actual AU annotations) or substantially weaken the claim that the PIN produces 'relatively objective' distributions.
  2. [III-D, Eq. (10)] Section III-D states that the Uncertainty Estimation Module assesses the consistency of predictions across N_sub subregions 'in TRN,' but Eq. (10) defines w^au_i using \tilde{p}_n, which are the PIN subregion soft predictions from Eq. (6), not TRN predictions. The text and the equation are therefore inconsistent. Please clarify whether the uncertainty weight is computed from PIN or TRN subregion predictions, and correct the description accordingly. This is not just a notation issue because it changes what quantity the confidence score measures.
  3. [IV-C, IV-G; Tables XI, XII; Fig. 12] The hyperparameters \lambda_1, \lambda_2, \lambda_3, T, and L_sub are selected based on RAF-DB test-set accuracy in the ablation study, yet the paper then reports the resulting configuration as the default for all datasets. The text in Section IV-C says each experiment is run six times to determine the average accuracy, but no standard deviations or significance tests are reported anywhere, and the ablations do not indicate whether a held-out validation split was used. Given that some benchmark margins are small (e.g., 0.16% on RAF-DB and 0.18% on AffectNet in Tables VI and VII), test-set hyperparameter tuning without error bars makes the claimed improvements difficult to assess. Please report standard deviations, use a validation split for hyperparameter selection, and state which comparisons are statistically significant.
minor comments (5)
  1. [Overall manuscript] Several blocks of text are duplicated in the manuscript: the paragraph beginning 'JOURNAL OF LATEX CLASS FILES...' and the full description of Section III-A appear twice, and the Fig. 2 caption is repeated. The manuscript should be cleaned up.
  2. [Eq. (4)] Equation (4) contains a typo: the second loss term reads '(1 − \hat{y}_{i,n,m}) log(1 − log \hat{p}_{i,n,m})', which should be '(1 − \hat{y}_{i,n,m}) log(1 − \hat{p}_{i,n,m})'.
  3. [Table V] In Table V, the bracket notation is inconsistent: POI* ResNet-18 is shown as '[99.55]' while POI ResNet-50 is shown as '99.89', and the text says 'the highest recognition accuracy (99.89%)' without clarifying that this comes from the ResNet-50 variant. Please explain the bracket convention and distinguish the configurations clearly.
  4. [IV-F] The comparison with SFEW 2.0 refers to 'Table XII', but the corresponding table is labeled Table IX. Please correct the cross-reference.
  5. [IV-B] The learning-rate schedule is described as 'drops by 10 after 20 epochs'; this should read 'drops by a factor of 10'.

Circularity Check

1 steps flagged · score 5.0 of 10

The 'objective' soft labels are not independent of the subjective annotation: Eq. (4) constructs AU pseudolabels from the emotion label y_i, so the PIN's outputs and the distilled soft labels inherit the very annotation bias they claim to correct.

  1. self definitional [Section III-C1 (Prior branches), Eq. (4) and surrounding text]
    "That is, if the i-th face image belongs to the c-th expression category, then we deploy the AU loss for each ˆpi,n,m: LAU = ... The pseudolabel is set according to Table I, AU pseudolabel ˆyi,n,m = 1 − ε when the AU categories represented by m-th AU net appear in the c-th emotional category, and 0 + ε otherwise. ε is used to suppress the influence of false AU pseudolabels under uncertain emotional annotations."

    All AU supervision in the PIN is generated by a deterministic lookup from the image's (possibly noisy) emotion label y_i through Table I; the image's actual AU pattern is never observed. Equations (4)-(8) therefore train the subregion predictions p_{i,n} and the intermediate 'objective' distribution ~p*_i on targets that are functions of y_i, and Eq. (11) distills those same label-derived estimates into the TRN. A wrong or biased y_i corrupts every AU pseudolabel for that sample in the same direction, so the claimed 'relatively objective' distribution inherits the annotation bias.

full rationale

The circularity is partial rather than total. The AU-expression correlation table (Table I, based on external works [27], [34]) is real prior knowledge, and the network still uses image features, so the empirical results are not vacuous. However, the paper's central claim that the PIN produces a 'relatively objective' emotion distribution that corrects subjective annotation ambiguity is not supported by an independent signal: the only channel through which the prior enters is Eq. (4), which maps each subjective label y_i to a fixed, smoothed AU pseudolabel vector. The entire objective-inference chain, from AU losses to the intermediate distribution and the TRN distillation target, is trained on targets constructed from the same labels the method claims to correct. The UEM confidence measure is similarly internal: it measures consistency among branches trained on the same label-derived targets, so it quantifies agreement with the label-derived prior rather than objectivity. The ablation 'w/o inpre' in Table III removes only the intermediate prediction and gate while retaining the same y_i-derived AU targets, so it cannot separate prior knowledge from label-derived regularization. Self-citations (e.g., [13], [16]) are present but not load-bearing, and the benchmark numbers may still reflect a useful noise-robust training recipe; the circularity concerns the 'objective inference' interpretation, not the accuracy figures themselves.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

POI's performance rests on hand-set hyperparameters (epsilon, T, Lsub, lambda) and on domain assumptions about facial anatomy and AU-emotion correlations. The AU prior is the only external knowledge; the soft labels themselves are produced by a network trained with the same labels, so the contribution is a training procedure rather than a new scientific quantity.

free parameters (4)
  • epsilon (AU pseudolabel smoothing) = 0.1
    Chosen by hand to suppress false AU pseudolabels under noisy emotion annotations; used in Eq. (4), Section III-C.1.
  • Temperature T = 3
    Tuned on RAF-DB test accuracy, Table XII; controls softness of intermediate predictions in Eq. (6).
  • Region size Lsub = 9
    Tuned on RAF-DB test accuracy, Table XI; controls crop size of facial subregions.
  • Loss weights lambda1, lambda2, lambda3 = 0.5, 0.5, 1.0
    Tuned via ablation on RAF-DB, Fig. 12; balance AU loss, PIN cross-entropy, and TRN cross-entropy.
assumptions (5)
  • domain assumption AU-expression correlation table (Table I) from [27], [34] is valid for in-the-wild facial images.
    Used as prior supervision for AU pseudolabels in Section III-C.1; if incorrect, the PIN's emotion inference is biased.
  • domain assumption Facial muscle movements relevant to emotion are localized near eyes and mouth, so cropping four corner subregions is sufficient.
    Section III-C.1, based on [47]-[49]; large poses or occlusions break this, as acknowledged in the failure examples.
  • domain assumption Left-right facial symmetry justifies flipping the right-side features to share structure across symmetrical subregions.
    Eq. (2), Section III-C.1.
  • ad hoc to paper Consistency across subregion predictions, measured as one minus the variance in Eq. (10), is a valid proxy for expression confidence.
    Section III-D, UEM; motivated by epistemic uncertainty in [50] but not derived, with Table IV used as supporting evidence.
  • domain assumption Random label flipping in synthetic experiments approximates real-world annotation ambiguity.
    Section IV-C; real ambiguity is structured and annotator-dependent, not random, so the simulation may not capture the target setting.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Prior-based Objective Inference Mining Potential Uncertainty for Facial Expression Recognition." pith.science (2026). https://pith.science/paper/PJ3FUFJU

@misc{pith2026241113024,
  author       = {Pith},
  title        = {Pith review of: Prior-based Objective Inference Mining Potential Uncertainty for Facial Expression Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PJ3FUFJU}},
  note         = {Machine review of arXiv:2411.13024}
}
read the original abstract

Annotation ambiguity caused by the inherent subjectivity of visual judgment has always been a major challenge for Facial Expression Recognition (FER) tasks, particularly for largescale datasets from in-the-wild scenarios. A potential solution is the evaluation of relatively objective emotional distributions to help mitigate the ambiguity of subjective annotations. To this end, this paper proposes a novel Prior-based Objective Inference (POI) network. This network employs prior knowledge to derive a more objective and varied emotional distribution and tackles the issue of subjective annotation ambiguity through dynamic knowledge transfer. POI comprises two key networks: Firstly, the Prior Inference Network (PIN) utilizes the prior knowledge of AUs and emotions to capture intricate motion details. To reduce over-reliance on priors and facilitate objective emotional inference, PIN aggregates inferential knowledge from various key facial subregions, encouraging mutual learning. Secondly, the Target Recognition Network (TRN) integrates subjective emotion annotations and objective inference soft labels provided by the PIN, fostering an understanding of inherent facial expression diversity, thus resolving annotation ambiguity. Moreover, we introduce an uncertainty estimation module to quantify and balance facial expression confidence. This module enables a flexible approach to dealing with the uncertainties of subjective annotations. Extensive experiments show that POI exhibits competitive performance on both synthetic noisy datasets and multiple real-world datasets. All codes and training logs will be publicly available at https://github.com/liuhw01/POI.

Figures

Figures reproduced from arXiv: 2411.13024 by the authors.

Figure 1
Figure 1. Subjective voting results and AUs judgment results for emotion [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Structure of proposed method. POI consists of a shared feature extractor, a Prior Inference Network, and a Target Recognition Network, which are g. 2. Structure of proposed method. POI consists of a shared feature extractor, a Prior Inference Network, and a Target Recognition Network, which are [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. AU groups with high correlation with emotion in key regions of left [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Expression consistency estimation for all facial subregions [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Partial sample of the dataset. The red box indicates that the sample [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 7
Figure 7. Figure 7: Inference results for mislabelled emotion images in prior inference [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 6
Figure 6. Figure 6: T-SNE visualizations of feature embeddings on the RAF-DB with [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: The emotional annotation categories and the potential emotion distribution derived by the PIN at different confidence scores. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 10
Figure 10. Figure 10: Visualization of AU detection results by the prior branch under RAF [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: Confusion matrices of (a) RAF-DB, (b) AffectNet, (c) FERPlus, and (d) SFEW 2.0 datasets. Su: Surprise, Fe: Fear, Di: Disgust, Ha: Happiness, Sa: [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 12
Figure 12. Figure 12: Impacts of weights λ1, λ2, and λ3 on performance of RAF-DB dataset. uncertain emotion annotations. The UEM further weights the confidence of facial expressions, enhancing the transmission of objective reasoning knowledge. 2) Influence of parameters: Region size Lsub: …
Figure 13
Figure 13. Figure 13: Accuracy with different fusion strategies of the auxiliary inference [PITH_FULL_IMAGE:figures/full_fig_p011_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

67 extracted references · 54 canonical work pages

  1. [1]

    Deep facial expression recognition: A survey,

    S. Li and W. Deng, “Deep facial expression recognition: A survey,” IEEE transactions on affective computing , 2020

    work page 2020

  2. [2]

    Multi-modal recurrent attention networks for facial expression recognition,

    J. Lee, S. Kim, S. Kim, and K. Sohn, “Multi-modal recurrent attention networks for facial expression recognition,”IEEE Transactions on Image Processing, vol. 29, pp. 6977–6991, 2020

    work page 2020

  3. [3]

    Facial expression recog- nition in videos using dynamic kernels,

    N. Perveen, D. Roy, and K. M. Chalavadi, “Facial expression recog- nition in videos using dynamic kernels,” IEEE Transactions on Image Processing, vol. 29, pp. 8316–8325, 2020

    work page 2020

  4. [4]

    Impact of deep learning approaches on facial expression recognition in healthcare industries,

    C. Bisogni, A. Castiglione, S. Hossain, F. Narducci, and S. Umer, “Impact of deep learning approaches on facial expression recognition in healthcare industries,” IEEE Transactions on Industrial Informatics , vol. 18, no. 8, pp. 5619–5627, 2022

    work page 2022

  5. [5]

    An architecture for emotional facial expressions as social signals,

    R. Aylett, C. Ritter, M. Y . Lim, F. Broz, P. E. McKenna, I. Keller, and G. Rajendran, “An architecture for emotional facial expressions as social signals,” IEEE Transactions on Affective Computing , vol. 12, no. 2, pp. 293–305, 2019

    work page 2019

  6. [6]

    The extended cohn-kanade dataset (ck+): A complete dataset for action unit and emotion-specified expression,

    P. Lucey, J. F. Cohn, T. Kanade, J. Saragih, Z. Ambadar, and I. Matthews, “The extended cohn-kanade dataset (ck+): A complete dataset for action unit and emotion-specified expression,” in 2010 ieee computer soci- ety conference on computer vision and pattern recognition-workshops . IEEE, 2010, pp. 94–101

    2010

  7. [7]

    Coding facial expressions with gabor wavelets,

    M. Lyons, S. Akamatsu, M. Kamachi, and J. Gyoba, “Coding facial expressions with gabor wavelets,” in Proceedings Third IEEE interna- tional conference on automatic face and gesture recognition . IEEE, 1998, pp. 200–205

    work page 1998

  8. [8]

    Training deep networks for facial expression recognition with crowd-sourced label distribution,

    E. Barsoum, C. Zhang, C. C. Ferrer, and Z. Zhang, “Training deep networks for facial expression recognition with crowd-sourced label distribution,” in Proceedings of the 18th ACM International Conference on Multimodal Interaction , 2016, pp. 279–283

    work page 2016

Show all 67 references
  1. [9]

    Affectnet: A database for facial expression, valence, and arousal computing in the wild,

    A. Mollahosseini, B. Hasani, and M. H. Mahoor, “Affectnet: A database for facial expression, valence, and arousal computing in the wild,” IEEE Transactions on Affective Computing , vol. 10, no. 1, pp. 18–31, 2017

    2017

  2. [10]

    Reliable crowdsourcing and deep locality- preserving learning for expression recognition in the wild,

    S. Li, W. Deng, and J. Du, “Reliable crowdsourcing and deep locality- preserving learning for expression recognition in the wild,” in Proceed- ings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 2852–2861

    2017

  3. [11]

    La-net: Landmark-aware learning for reliable facial expression recognition under label noise,

    Z. Wu and J. Cui, “La-net: Landmark-aware learning for reliable facial expression recognition under label noise,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 20 698–20 707

    2023

  4. [12]

    Relative uncertainty learning for fa- cial expression recognition,

    Y . Zhang, C. Wang, and W. Deng, “Relative uncertainty learning for fa- cial expression recognition,” Advances in Neural Information Processing Systems, vol. 34, pp. 17 616–17 627, 2021

    2021

  5. [13]

    Learning from more: Combating uncertainty cross-multidomain for facial expression recog- nition,

    H. Liu, H. Cai, Q. Lin, X. Li, and H. Xiao, “Learning from more: Combating uncertainty cross-multidomain for facial expression recog- nition,” in Proceedings of the 31st ACM International Conference on Multimedia, 2023, pp. 5889–5898

    2023

  6. [14]

    Discovering cultural differences (and similar- ities) in facial expressions of emotion,

    C. Chen and R. E. Jack, “Discovering cultural differences (and similar- ities) in facial expressions of emotion,” Current opinion in psychology , vol. 17, pp. 61–66, 2017

    2017

  7. [15]

    Cultural confusions show that facial expressions are not universal,

    R. E. Jack, C. Blais, C. Scheepers, P. G. Schyns, and R. Caldara, “Cultural confusions show that facial expressions are not universal,” Current biology, vol. 19, no. 18, pp. 1543–1548, 2009

    2009

  8. [16]

    Adaptive multilayer perceptual attention network for facial expression recognition,

    H. Liu, H. Cai, Q. Lin, X. Li, and H. Xiao, “Adaptive multilayer perceptual attention network for facial expression recognition,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 9, pp. 6253–6266, 2022

    2022

  9. [17]

    Dynamic facial expression recognition under partial occlusion with optical flow reconstruction,

    D. Poux, B. Allaert, N. Ihaddadene, I. M. Bilasco, C. Djeraba, and M. Bennamoun, “Dynamic facial expression recognition under partial occlusion with optical flow reconstruction,” IEEE Transactions on Image Processing, vol. 31, pp. 446–457, 2021

    2021

  10. [18]

    Compound facial expressions of emotion,

    S. Du, Y . Tao, and A. M. Martinez, “Compound facial expressions of emotion,” Proceedings of the national academy of sciences , vol. 111, no. 15, pp. E1454–E1462, 2014

    2014

  11. [19]

    Suppressing uncer- tainties for large-scale facial expression recognition,

    K. Wang, X. Peng, J. Yang, S. Lu, and Y . Qiao, “Suppressing uncer- tainties for large-scale facial expression recognition,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 6897–6906

    2020

  12. [20]

    Facial emotion recognition with noisy multi-task annotations,

    S. Zhang, Z. Huang, D. P. Paudel, and L. Van Gool, “Facial emotion recognition with noisy multi-task annotations,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , 2021, pp. 21–31

    2021

  13. [21]

    Dive into ambi- guity: Latent distribution mining and pairwise uncertainty estimation for facial expression recognition,

    J. She, Y . Hu, H. Shi, J. Wang, Q. Shen, and T. Mei, “Dive into ambi- guity: Latent distribution mining and pairwise uncertainty estimation for facial expression recognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2021, pp. 6248–6257

    2021

  14. [22]

    Towards facial expression recognition in the wild via noise-tolerant network,

    Y . Gu, H. Yan, X. Zhang, Y . Wang, Y . Ji, and F. Ren, “Towards facial expression recognition in the wild via noise-tolerant network,” IEEE Transactions on Circuits and Systems for Video Technology , 2022

    2022

  15. [23]

    Label distribution learning on auxiliary label space graphs for facial expression recognition,

    S. Chen, J. Wang, Y . Chen, Z. Shi, X. Geng, and Y . Rui, “Label distribution learning on auxiliary label space graphs for facial expression recognition,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 13 984–13 993

    2020

  16. [24]

    Bias-based soft label learn- ing for facial expression recognition,

    S. Wang, H. Shuai, C. Liu, and Q. Liu, “Bias-based soft label learn- ing for facial expression recognition,” IEEE Transactions on Affective Computing, 2022

    2022

  17. [25]

    A unified deep model for joint facial expression recognition, face synthesis, and face alignment,

    F. Zhang, T. Zhang, Q. Mao, and C. Xu, “A unified deep model for joint facial expression recognition, face synthesis, and face alignment,” IEEE Transactions on Image Processing , vol. 29, pp. 6574–6589, 2020

    2020

  18. [26]

    Automatic analysis of facial actions: A survey,

    B. Martinez, M. F. Valstar, B. Jiang, and M. Pantic, “Automatic analysis of facial actions: A survey,” IEEE transactions on affective computing , vol. 10, no. 3, pp. 325–347, 2017

    2017

  19. [27]

    Exploring domain knowledge for facial expression-assisted action unit activation recognition,

    S. Wang, G. Peng, and Q. Ji, “Exploring domain knowledge for facial expression-assisted action unit activation recognition,” IEEE Transac- tions on Affective Computing , vol. 11, no. 4, pp. 640–652, 2018

    2018

  20. [28]

    Fine-grained facial expression recognition in the wild,

    L. Liang, C. Lang, Y . Li, S. Feng, and J. Zhao, “Fine-grained facial expression recognition in the wild,” IEEE Transactions on Information Forensics and Security, vol. 16, pp. 482–494, 2020

    2020

  21. [29]

    Prior aided streaming network for multi-task affective analysis,

    W. Zhang, Z. Guo, K. Chen, L. Li, Z. Zhang, Y . Ding, R. Wu, T. Lv, and C. Fan, “Prior aided streaming network for multi-task affective analysis,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 3539–3549

    2021

  22. [30]

    Using positive matching contrastive loss with facial action units to mitigate bias in facial expression recognition,

    V . Suresh and D. C. Ong, “Using positive matching contrastive loss with facial action units to mitigate bias in facial expression recognition,” in 2022 10th International Conference on Affective Computing and Intelligent Interaction (ACII) . IEEE, 2022, pp. 1–8

    2022

  23. [31]

    Micro-expression recognition based on facial graph representation learning and facial action unit fusion,

    L. Lei, T. Chen, S. Li, and J. Li, “Micro-expression recognition based on facial graph representation learning and facial action unit fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 1571–1580

    2021

  24. [32]

    Au-expression knowledge constrained representation learning for facial expression recognition,

    T. Pu, T. Chen, Y . Xie, H. Wu, and L. Lin, “Au-expression knowledge constrained representation learning for facial expression recognition,” in 2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 11 154–11 161

    2021

  25. [33]

    Facial expression recognition via deep action units graph network based on psychological mechanism,

    Y . Liu, X. Zhang, Y . Lin, and H. Wang, “Facial expression recognition via deep action units graph network based on psychological mechanism,” IEEE Transactions on Cognitive and Developmental Systems , vol. 12, no. 2, pp. 311–322, 2019

    2019

  26. [34]

    Knowledge augmented deep neural networks for joint facial expression and action unit recognition,

    Z. Cui, T. Song, Y . Wang, and Q. Ji, “Knowledge augmented deep neural networks for joint facial expression and action unit recognition,” Ad- vances in Neural Information Processing Systems , vol. 33, pp. 14 338– 14 349, 2020

    2020

  27. [35]

    Objective class-based micro-expression recognition through simultaneous action unit detection and feature aggregation,

    L. Zhou, Q. Mao, and M. Dong, “Objective class-based micro-expression recognition through simultaneous action unit detection and feature aggregation,” arXiv preprint arXiv:2012.13148 , 2020

    2012 arXiv

  28. [36]

    Weakly supervised facial action unit recognition through adversarial training,

    G. Peng and S. Wang, “Weakly supervised facial action unit recognition through adversarial training,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 2188–2196

    2018

  29. [37]

    Facial expression recognition with visual transformers and attentional selective fusion,

    F. Ma, B. Sun, and S. Li, “Facial expression recognition with visual transformers and attentional selective fusion,” IEEE Transactions on Affective Computing, 2021

    2021

  30. [38]

    Region attention networks for pose and occlusion robust facial expression recognition,

    K. Wang, X. Peng, J. Yang, D. Meng, and Y . Qiao, “Region attention networks for pose and occlusion robust facial expression recognition,” IEEE Transactions on Image Processing, vol. 29, pp. 4057–4069, 2020

    2020

  31. [39]

    Joint expression synthesis and repre- sentation learning for facial expression recognition,

    X. Zhang, F. Zhang, and C. Xu, “Joint expression synthesis and repre- sentation learning for facial expression recognition,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 32, no. 3, pp. 1681– 1695, 2021

    2021

  32. [40]

    D 2 s: Dynamic distribution supervision for multi-label facial expression recognition,

    L. Wang, X. Zhang, N. Jiang, H. Wu, and J. Yang, “D 2 s: Dynamic distribution supervision for multi-label facial expression recognition,” in 2022 IEEE International Conference on Multimedia and Expo (ICME) . IEEE, 2022, pp. 1–6

    2022

  33. [41]

    Facial action coding system,

    P. Ekman and W. V . Friesen, “Facial action coding system,” Environ- mental Psychology & Nonverbal Behavior , 1978. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2023 13

    1978

  34. [42]

    Relational knowledge distilla- tion,

    W. Park, D. Kim, Y . Lu, and M. Cho, “Relational knowledge distilla- tion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 3967–3976

    2019

  35. [43]

    Deep mutual learning,

    Y . Zhang, T. Xiang, T. M. Hospedales, and H. Lu, “Deep mutual learning,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 4320–4328

    2018

  36. [44]

    Collaborative learning for deep neural networks,

    G. Song and W. Chai, “Collaborative learning for deep neural networks,” Advances in neural information processing systems , vol. 31, 2018

    2018

  37. [45]

    Cooperative learning with visual attributes,

    T. Batra and D. Parikh, “Cooperative learning with visual attributes,” arXiv preprint arXiv:1705.05512 , 2017

    2017 arXiv

  38. [46]

    Learning deep global multi-scale and local attention features for facial expression recognition in the wild,

    Z. Zhao, Q. Liu, and S. Wang, “Learning deep global multi-scale and local attention features for facial expression recognition in the wild,” IEEE Transactions on Image Processing, vol. 30, pp. 6544–6556, 2021

    2021

  39. [47]

    Orientations for the successful categorization of facial expressions and their link with facial features,

    J. Duncan, F. Gosselin, C. Cobarro, G. Dugas, C. Blais, and D. Fiset, “Orientations for the successful categorization of facial expressions and their link with facial features,” Journal of vision , vol. 17, no. 14, pp. 7–7, 2017

    2017

  40. [48]

    Transfer: Learning relation-aware facial expression representations with transformers,

    F. Xue, Q. Wang, and G. Guo, “Transfer: Learning relation-aware facial expression representations with transformers,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 3601–3610

    2021

  41. [49]

    Adaptive deep disturbance-disentangled learning for facial expression recogni- tion,

    D. Ruan, R. Mo, Y . Yan, S. Chen, J.-H. Xue, and H. Wang, “Adaptive deep disturbance-disentangled learning for facial expression recogni- tion,” International Journal of Computer Vision , vol. 130, no. 2, pp. 455–477, 2022

    2022

  42. [50]

    What uncertainties do we need in bayesian deep learning for computer vision?

    A. Kendall and Y . Gal, “What uncertainties do we need in bayesian deep learning for computer vision?” Advances in neural information processing systems, vol. 30, 2017

    2017

  43. [51]

    Collecting large, richly annotated facial-expression databases from movies,

    A. Dhall, R. Goecke, S. Lucey, and T. Gedeon, “Collecting large, richly annotated facial-expression databases from movies,” IEEE multimedia , vol. 19, no. 03, pp. 34–41, 2012

    2012

  44. [52]

    Vggface2: A dataset for recognising faces across pose and age,

    Q. Cao, L. Shen, W. Xie, O. M. Parkhi, and A. Zisserman, “Vggface2: A dataset for recognising faces across pose and age,” in 2018 13th IEEE international conference on automatic face & gesture recognition (FG 2018). IEEE, 2018, pp. 67–74

    2018

  45. [53]

    Retinaface: Single-shot multi-level face localisation in the wild,

    J. Deng, J. Guo, E. Ververas, I. Kotsia, and S. Zafeiriou, “Retinaface: Single-shot multi-level face localisation in the wild,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 5203–5212

    2020

  46. [54]

    Mitigating label-noise for facial expression recognition in the wild,

    H. Yan, Y . Gu, X. Zhang, Y . Wang, Y . Ji, and F. Ren, “Mitigating label-noise for facial expression recognition in the wild,” in 2022 IEEE International Conference on Multimedia and Expo (ICME) . IEEE, 2022, pp. 1–6

    2022

  47. [55]

    Co-attentive multi-task convolutional neural network for facial expression recognition,

    W. Yu and H. Xu, “Co-attentive multi-task convolutional neural network for facial expression recognition,” Pattern Recognition , vol. 123, p. 108401, 2022

    2022

  48. [56]

    Triplet loss with multistage outlier suppression and class-pair margins for facial expres- sion recognition,

    W. Xie, H. Wu, Y . Tian, M. Bai, and L. Shen, “Triplet loss with multistage outlier suppression and class-pair margins for facial expres- sion recognition,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 32, no. 2, pp. 690–703, 2021

    2021

  49. [57]

    Convolution by multiplication: Accelerated two-stream fourier domain convolutional neural network for facial expression recognition,

    M. Huang, X. Zhang, X. Lan, H. Wang, and Y . Tang, “Convolution by multiplication: Accelerated two-stream fourier domain convolutional neural network for facial expression recognition,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 3, pp. 1431– ...

    2021

  50. [58]

    Dual-channel feature disentanglement for identity-invariant facial expression recognition,

    Y . Li, Y . Lu, M. Gong, L. Liu, and L. Zhao, “Dual-channel feature disentanglement for identity-invariant facial expression recognition,” Information Sciences, vol. 608, pp. 410–423, 2022

    2022

  51. [59]

    Self-paced label distribution learning for in-the-wild facial expression recognition,

    J. Shao, Z. Wu, Y . Luo, S. Huang, X. Pu, and Y . Ren, “Self-paced label distribution learning for in-the-wild facial expression recognition,” in Proceedings of the 30th ACM International Conference on Multimedia , 2022, pp. 161–169

    2022

  52. [60]

    Weakly-supervised facial expression recognition in the wild with noisy data,

    F. Zhang, M. Xu, and C. Xu, “Weakly-supervised facial expression recognition in the wild with noisy data,” IEEE Transactions on Mul- timedia, vol. 24, pp. 1800–1814, 2021

    2021

  53. [61]

    Proba- bilistic attribute tree structured convolutional neural networks for facial expression recognition in the wild,

    J. Cai, Z. Meng, A. S. Khan, Z. Li, J. O’Reilly, and Y . Tong, “Proba- bilistic attribute tree structured convolutional neural networks for facial expression recognition in the wild,” IEEE Transactions on Affective Computing, 2022

    2022

  54. [62]

    Learn from all: Erasing attention consistency for noisy label facial expression recognition,

    Y . Zhang, C. Wang, X. Ling, and W. Deng, “Learn from all: Erasing attention consistency for noisy label facial expression recognition,” in European Conference on Computer Vision . Springer, 2022, pp. 418– 434

    2022

  55. [63]

    Face2exp: Com- bating data biases for facial expression recognition,

    D. Zeng, Z. Lin, X. Yan, Y . Liu, F. Wang, and B. Tang, “Face2exp: Com- bating data biases for facial expression recognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 20 291–20 300

    2022

  56. [64]

    Region attention enhanced unsu- pervised cross-domain facial emotion recognition,

    Y . Ji, Y . Hu, Y . Yang, and H. T. Shen, “Region attention enhanced unsu- pervised cross-domain facial emotion recognition,” IEEE Transactions on Knowledge and Data Engineering , 2021

    2021

  57. [65]

    Discriminant distribution-agnostic loss for facial expression recognition in the wild,

    A. H. Farzaneh and X. Qi, “Discriminant distribution-agnostic loss for facial expression recognition in the wild,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020, pp. 406–407

    2020

  58. [66]

    Boosting facial expression recognition by a semi-supervised progressive teacher,

    J. Jiang and W. Deng, “Boosting facial expression recognition by a semi-supervised progressive teacher,” IEEE Transactions on Affective Computing, 2021

    2021

  59. [67]

    Robust facial expression recognition with convolutional visual transformers,

    F. Ma, B. Sun, and S. Li, “Robust facial expression recognition with convolutional visual transformers,” arXiv preprint arXiv:2103.16854 , 2021

    2021 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.