Pith. sign in

REVIEW 3 major objections 5 minor 35 references

Lifting Scheme-Based Implicit Disentanglement of Emotion-Related Facial Dynamics in the Wild

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read The paper proposes IFDD, a fully learnable wavelet-lifting framework that implicitly separates emotion-related facial dynamics from emotion-irrelevant global context in video, and reports that it outperforms prior supervised dynamic…

desk verdict Real and sizable accuracy gains on three DFER benchmarks, but the 'implicit disentanglement' claim is more interpretation than demonstrated property. read the letter →

arxiv 2412.13168 v2 pith:MOB2N5AH submitted 2024-12-17 cs.CV cs.AI

classification cs.CVcs.AI
keywords dynamicfacialexpressionrecognitionwaveletliftingschemeimplicitdisentanglementspatiotemporalfeaturesplittingglobalcontextin-the-wildvideocross-attentionvisiontransformer
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

In-the-wild expression videos are mostly neutral or noisy frames, so emotion-relevant motion is diluted by global context and identity-bound appearance. This paper claims that the right way to handle that dilution is not to hand-design a motion detector—differencing frames, frequency analysis, optical flow—but to let a network learn to split video features into a static global-context group and a dynamic group, then refine the split with a learnable version of the wavelet lifting scheme. The result, IFDD, is claimed to outperform previous supervised methods on DFEW, FERV39k, and MAFW, with the largest reported gain on DFEW (73.82% WAR versus 69.25% for the best prior supervised model) and only modest added compute. If right, this gives a recipe, needing no external guidance or explicit motion priors, for getting cleaner expression representations out of raw video.

What carries the argument

The load-bearing object is the wavelet lifting scheme (split, predict, update), expanded into a fully learnable, two-stage module. Standard lifting splits a signal into even and odd samples before predicting and updating; IFDD replaces the fixed split with ISSM, which uses softmax attention over temporal tokens to generate content-aware interpolated indices, and replaces the fixed predictor and updater with cross-attention MLP blocks in LADM, where the updater aggregates both groups and the predictor strips global context from the dynamic group. The resulting division is $Y_S$ for low-frequency global context and $Y_D$ for the residual high-frequency emotion dynamics, and the classification loss is applied only to $Y_D$.

What would settle it

Synthesize a benchmark in which the same identity, lighting, and background are held fixed while only the labeled expression changes, and train IFDD on it; then check whether the dynamic branch $Y_D$ is the only branch whose linear probe predicts the label. If the static branch $Y_S$ also recovers the emotion label, or if a label-shuffled version of the data still allows $Y_D$ to perform well, the disentanglement claim would be refuted.

Watch

Extended reading notes

Core claim

The central claim is that emotion-related dynamics can be disentangled from global context implicitly, by treating the video feature tensor as a signal and applying an expanded lifting scheme. ISSM learns content-aware temporal splitting indices from inter-frame correlation and interpolates the frame features into two halves: a relatively static group and a relatively dynamic group. LADM then runs an updater that aggregates both groups into a refined global-context estimate $Y_S$, and a predictor that subtracts global-context content from the dynamic group to yield purified emotion dynamics $Y_D$; a global-context loss forces $Y_S$ to keep the same local average as the input along temporal and channel dimensions, so the static branch is pinned to the global appearance. The classification head is applied only to $Y_D$. On DFEW, FERV39k, and MAFW the paper reports that IFDD variants surpass prior supervised DFER methods, and ablations tie the gains to both modules and to the global-context loss.

Load-bearing premise

The whole argument hinges on the global-context loss: forcing the static branch $Y_S$ to match the local temporal-and-channel average of the input is what supposedly makes it absorb identity, background, and other emotion-irrelevant context, leaving $Y_D$ as purified emotion dynamics; if that constraint instead discards useful emotion cues or fails to absorb irrelevant context, the reported accuracy gains do not establish genuine disentanglement.

Editorial extensions

If this is right

  • On the three in-the-wild benchmarks, IFDD-3DViT reports WAR of 73.82% on DFEW, 51.09% on FERV39k, and 53.92% on MAFW, beating the best prior supervised methods compared in the paper.
  • Ablations show the two stages are complementary: ISSM alone and LADM alone each improve over the backbone, and using both gives the largest gain.
  • The global-context loss is load-bearing: removing it drops UAR by 2.63 percentage points on DFEW, and constraining local averages along $\{T,C\}$ performs better than constraining all four dimensions.
  • IFDD is backbone-agnostic: the same modules attach to a 2D CNN (MobileNetV2) and a 3D ViT (MViT-S), with modest extra FLOPs of about 3.5% and 9.7%.
  • Because classification uses only $Y_D$, the framework predicts from residual emotion dynamics rather than from the full coupled spatiotemporal representation.

Reading between the lines

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

  • If the disentanglement generalizes, the same lifting recipe could apply to other dynamic tasks the authors mention—micro-expression recognition, optical flow estimation, and video compression—where separating static context from temporal residual is the core difficulty.
  • The implicit-disentanglement claim suggests a testable signature: the static branch $Y_S$ should carry identity and background information while being nearly label-blind, and the dynamic branch should be the opposite; a reader could verify this by training linear probes on each branch.
  • Because the splitting indices are themselves learned from temporal correlation, the method could in principle adapt to variable-length or irregularly sampled video, a setting the current fixed-length-clip experiments do not cover.
  • The comparison excludes self-supervised methods pretrained on large external data; the paper notes MAE-DFER still exceeds IFDD-3DViT on DFEW, so the practical claim is specifically about supervised training without external sources.
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 IFDD, a wavelet-lifting-inspired framework for dynamic facial expression recognition (DFER). IFDD consists of an Inter-frame Static-dynamic Splitting Module (ISSM), which adaptively splits latent frame features into static and dynamic groups, and a Lifting-based Aggregation-Disentanglement Module (LADM), which refines these groups with cross-attention updaters and predictors. The final dynamic features are used for classification, and a global context loss encourages the static branch to match a temporally pooled local average of the input. The authors evaluate IFDD with MobileNetV2 and MViT-S backbones on DFEW, FERV39k, and MAFW, reporting consistent gains over prior supervised methods, with per-emotion tables, ablations, and efficiency comparisons included.

Significance. If the central claim is correct, the paper offers a genuinely useful architectural pattern: a lightweight, backbone-agnostic module that removes emotion-irrelevant global context while keeping emotion-related dynamics, at modest computational overhead. The strengths are the breadth of experiments (three in-the-wild datasets, two backbones, many ablations, per-emotion accuracy, efficiency tables), the clear writing of the framework, the release of code, and the fair decision to compare only against supervised methods in the main tables. The main weakness is that the specific disentanglement claim is not directly tested; the evidence is indirect, consisting of downstream classification accuracy and qualitative visualizations. Because the accuracy gains could in principle arise from added capacity or feature refinement rather than from principled context removal, the central conceptual claim needs additional quantitative verification before the paper can be fully accepted.

major comments (3)
  1. [Decoupling Loss, Eq. (9)] The objective does not actually enforce disentanglement between YD and YS. LLift only requires that YS has the same local average as X along the temporal and channel dimensions; there is no term that penalizes emotion-irrelevant, identity, or background content in YD, and no orthogonality or independence penalty between YD and YS. A degenerate solution is not excluded: the predictor P could output zero, leaving YD = XD, while the updater U learns to make YS match the pooled average of X. In that case the same losses would be minimized as long as classification on XD performs well, and the reported accuracy gains would not validate the specific claim of implicit disentanglement. I ask for an independent, quantitative test of what YD actually removes, for example by training linear probes for identity or neutral/background content on YD, by measuring mutual information or distance correlation between YD and YS, or by using a synthetic setting where contamination is known and the model's ability to remove it can be directly verified. The Grad-CAM and t-SNE analyses in Section 'Visualization Analysis' are qualitative and do not resolve this issue.
  2. [Ablation Study, Table 1] For the 3DViT backbone, the ISSM-only variant achieves UAR 55.53, which is lower than the 3DViT baseline's UAR of 57.00—a drop of 1.47 percentage points—even though WAR improves from 68.60 to 71.68. Since UAR is one of the two primary metrics and the text states that 'ISSM and LADM both show remarkable effectiveness in recognition accuracy,' the claim is only partially supported. The interaction between ISSM and the 3DViT backbone needs explanation, or the claim should be softened to reflect that ISSM alone is not consistently beneficial across backbones and metrics.
  3. [Implementation Details; Appendix F] There is a direct contradiction about how the DFEW and MAFW numbers are obtained. The main text says 'As for DFER and MAFW with 5-fold cross-validation, we report metrics from a single run,' while Appendix F says that the final results are the average of 5-fold results and Table 13 lists per-fold numbers. If the main results are from a single run, then all comparisons on DFEW and MAFW have no error bars, which weakens the claimed margins over prior methods. If they are 5-fold averages, the text should say so and should provide fold-wise standard deviations or at least the variance across folds for the main tables. Please clarify and report run-to-run variability for the headline numbers.
minor comments (5)
  1. [Eq. (4)] The formula for XD[i] appears to use the index variable IF in some places and IS in others; this is likely a typo and should be made consistent with the definition of ID.
  2. [Experiments, Datasets and Metrics] The phrase 'As for DFER and MAFW' should read 'As for DFEW and MAFW'; also the text near 'which all include neural emotion' should say 'neutral emotion.'
  3. [Visualization Analysis] The sentence 'We further visualize the gradient attention of YF and YS' uses YF where YD is intended; please correct the symbol.
  4. [Table 5, Freq-HD row] The Freq-HD row in Table 5 reports per-emotion numbers such as 6.9 and 64.61 that are inconsistent with the stated UAR/WAR and with the other rows; please verify the values and formatting.
  5. [Abstract and Introduction] The phrase 'without exploit operations and external guidance' should likely be 'without explicit operations and external guidance'; the grammar should be corrected throughout.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: held-out benchmark accuracy is self-contained; the single self-citation (Tang and Chai 2024) is non-load-bearing; the 'implicit disentanglement' label is an interpretive framing corroborated by independent visualizations, not a conclusion forced by the equations.

full rationale

Central claims are empirical, self-contained benchmark results. IFDD-3DViT/2DCNN accuracies are reported on held-out folds of DFEW (5-fold CV), FERV39k (train/test split), and MAFW (5-fold CV) under the same protocols as the compared supervised methods; ablations in Tables 1-4 vary ISSM/LADM components, splitting dependency/manner, mutual relation, and constrained dims of the global-context loss, all on the same held-out folds. No fitted parameter is renamed as a prediction: hyperparameters (allowable range L, token dim dT) are selected on DFEW validation partitions and reported openly (Table 9). The only self-reference, (Tang and Chai 2024), appears in the Introduction as a cited example of external-guidance methods ('without external guidance such as optical flows(Tang and Chai 2024) or facial landmarks(Luo et al. 2023)'); it is non-load-bearing and does not justify any premise of the method, so it does not raise the circularity score. The paper also honestly scopes its comparative claim to supervised methods, conceding in Appendix G that MAE-DFER surpasses IFDD-3DViT on DFEW, and it reports single-run 5-fold results rather than cherry-picked runs. The residual interpretive step is not a circular reduction: the headline 'implicit disentanglement' is attached to the lifting-style construction YD = XD - P(YS|XD) and YS = XS + U(XD|XS) (Eq. 5), regularized only by LLift = Huber(Avg_{T,C}(YS - PT_down2(X))) (Eq. 9) plus LCLS on YD. Because no term enforces that YS be emotion-free or that YD and YS be independent, the labels 'emotion-related' vs 'emotion-irrelevant' are empirical hypotheses corroborated by held-out accuracy, Grad-CAM (Fig. 2a), and t-SNE (Fig. 2b) -- not conclusions forced by the equations. That under-determination is a validity/correctness caveat rather than circularity, since no equation in the paper reduces a reported prediction to its own inputs by construction.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim rests on two domain assumptions about what constitutes global context and how to separate it, plus a small number of hand-set hyperparameters. The wavelet lifting mathematics is standard. No new physical or conceptual entities are introduced.

free parameters (3)
  • allowable_range_L = +-T/2 (T=8 for IFDD)
    Hyperparameter controlling the maximum offset of splitting indices in Eq. 3; ablation in Table 9 selects +-T/2 over +-T/4 and midpoint initialization.
  • global_context_loss_weight = 1 (implicit)
    The total loss L in Eq. 8 sums LCLS and LLif t with equal weight; no sensitivity analysis is reported for this weighting.
  • temporal_embedding_dim_dT = 128
    Set for both backbones in implementation details; no ablation reported for this dimension.
assumptions (3)
  • standard math The wavelet lifting scheme as described by Sweldens decomposes a signal into low- and high-frequency sub-bands via split, predict, and update steps.
    Used as the conceptual basis for ISSM and LADM in the sections on Inter-Frame Static-Dynamic Splitting and Lifting-based Aggregation and Disentanglement.
  • domain assumption Frames that are temporally correlated with all other frames in a clip constitute emotion-irrelevant global context, while less correlated frames carry emotion-related dynamics.
    Central to ISSM's design in Eq. 2 and the interpretation of XS and XD; no direct evidence is provided beyond downstream accuracy.
  • domain assumption Enforcing YS to match the local average of X along the {T, C} dimensions separates global context from dynamics.
    The global context loss in Eq. 9 instantiates this; if false, YD may still contain context or lose emotion information.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Lifting Scheme-Based Implicit Disentanglement of Emotion-Related Facial Dynamics in the Wild." pith.science (2026). https://pith.science/paper/MOB2N5AH

@misc{pith2026241213168,
  author       = {Pith},
  title        = {Pith review of: Lifting Scheme-Based Implicit Disentanglement of Emotion-Related Facial Dynamics in the Wild},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MOB2N5AH}},
  note         = {Machine review of arXiv:2412.13168}
}
read the original abstract

In-the-wild dynamic facial expression recognition (DFER) encounters a significant challenge in recognizing emotion-related expressions, which are often temporally and spatially diluted by emotion-irrelevant expressions and global context. Most prior DFER methods directly utilize coupled spatiotemporal representations that may incorporate weakly relevant features with emotion-irrelevant context bias. Several DFER methods highlight dynamic information for DFER, but following explicit guidance that may be vulnerable to irrelevant motion. In this paper, we propose a novel Implicit Facial Dynamics Disentanglement framework (IFDD). Through expanding wavelet lifting scheme to fully learnable framework, IFDD disentangles emotion-related dynamic information from emotion-irrelevant global context in an implicit manner, i.e., without exploit operations and external guidance. The disentanglement process contains two stages. The first is Inter-frame Static-dynamic Splitting Module (ISSM) for rough disentanglement estimation, which explores inter-frame correlation to generate content-aware splitting indexes on-the-fly. We utilize these indexes to split frame features into two groups, one with greater global similarity, and the other with more unique dynamic features. The second stage is Lifting-based Aggregation-Disentanglement Module (LADM) for further refinement. LADM first aggregates two groups of features from ISSM to obtain fine-grained global context features by an updater, and then disentangles emotion-related facial dynamic features from the global context by a predictor. Extensive experiments on in-the-wild datasets have demonstrated that IFDD outperforms prior supervised DFER methods with higher recognition accuracy and comparable efficiency. Code is available at https://github.com/CyberPegasus/IFDD.

Figures

Figures reproduced from arXiv: 2412.13168 by the authors.

Figure 1
Figure 1. Overview framework of IFDD, which mainly consists of four parts: (1) multiscale backbone followed by pyra [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Visualization analysis on the gradient attention of [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Schematic diagram of different ISSM variants. T, [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visual Comparison with baselines and prior meth [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Distribution of predicted per-class confidence on [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 29 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    X.; Gruson, A.; Polania, L

    Bastidas Rodriguez, M. X.; Gruson, A.; Polania, L. F.; Fujieda, S.; Ortiz, F. P.; Takayama, K.; and Hachisuka, T. 2020. Deep Adaptive Wavelet Network . In IEEE Winter Conf. Appl. Comput. Vis., 3100--3108

  4. [4]

    Carreira, J.; and Zisserman, A. 2017. Quo Vadis , Action Recognition ? A New Model and the Kinetics Dataset . In IEEE Conf. Comput. Vis. Pattern Recog., 4724--4733

  5. [5]

    Chen, D.; Wen, G.; Li, H.; Yang, P.; Chen, C.; and Wang, B. 2024. CDGT: Constructing diverse graph transformers for emotion recognition from facial videos. Neural Networks, 179: 106573

  6. [6]

    S.; Keltner, D.; Schroff, F.; Jou, B.; Adam, H.; and Prasad, G

    Cowen, A. S.; Keltner, D.; Schroff, F.; Jou, B.; Adam, H.; and Prasad, G. 2021. Sixteen Facial Expressions Occur in Similar Contexts Worldwide . Nature, 589(7841): 251--257

  7. [7]

    Deng, X.; Gao, C.; and Xu, M. 2023. PIRNet : Privacy-Preserving Image Restoration Network via Wavelet Lifting . In Int. Conf. Comput. Vis., 22311--22320

  8. [8]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; Uszkoreit, J.; and Houlsby, N. 2021. An Image is Worth 16x16 Words : Transformers for Image Recognition at Scale . arXiv:2010.11929

Show all 35 references
  1. [9]

    Fan, H.; Xiong, B.; Mangalam, K.; Li, Y.; Yan, Z.; Malik, J.; and Feichtenhofer, C. 2021. Multiscale Vision Transformers . In Int. Conf. Comput. Vis., 6804--6815

  2. [10]

    Hara, K.; Kataoka, H.; ; and Satoh, Y. 2018. Can Spatiotemporal 3d Cnns Retrace the History of 2d Cnns and Imagenet? In IEEE Conf. Comput. Vis. Pattern Recog., 6546--6555

  3. [11]

    Huang, H.; and Fang, Y. 2021. Adaptive Wavelet Transformer Network for 3D Shape Representation Learning . In Int. Conf. Learn. Represent

  4. [12]

    Huang, J.-J.; and Dragotti, P. L. 2022. WINNet: Wavelet-Inspired Invertible Network for Image Denoising . IEEE Trans. Image Process., 31: 4377--4392

  5. [13]

    Jiang, X.; Zong, Y.; Zheng, W.; Tang, C.; Xia, W.; Lu, C.; and Liu, J. 2020. DFEW : A Large-Scale Database for Recognizing Dynamic Facial Expressions in the Wild . In ACM Int. Conf. Multimedia, 2881--2889

  6. [14]

    Li, H.; Niu, H.; Zhu, Z.; and Zhao, F. 2023. Intensity- Aware Loss for Dynamic Facial Expression Recognition in the Wild . AAAI Conf. Artif. Intell., 37(1): 67--75

  7. [15]

    Li, H.; Sui, M.; Zhu, Z.; and zhao , F. 2022. NR-DFERNet : Noise-Robust Network for Dynamic Facial Expression Recognition . arXiv:2206.04975

  8. [16]

    Liu, Y.; Dai, W.; Feng, C.; Wang, W.; Yin, G.; Zeng, J.; and Shan, S. 2022. MAFW : A Large-scale , Multi-modal , Compound Affective Database for Dynamic Facial Expression Recognition in the Wild . In ACM Int. Conf. Multimedia, 24--32

  9. [17]

    Luo, B.; Zhu, J.; Yang, T.; Zhao, S.; Hu, C.; Zhao, X.; and Gao, Y. 2023. Learning Deep Hierarchical Features with Spatial Regularization for One-Class Facial Expression Recognition . AAAI Conf. Artif. Intell., 37(5): 6065--6073

  10. [18]

    Ma, F.; Sun, B.; and Li, S. 2022. Spatio- Temporal Transformer for Dynamic Facial Expression Recognition in the Wild . arXiv:2205.04749

  11. [19]

    Qiu, Z.; Yao, T.; and Mei, T. 2017. Learning Spatio-Temporal Representation with Pseudo-3D Residual Networks . In Int. Conf. Comput. Vis., 5534--5542

  12. [20]

    Sun, L.; Lian, Z.; Liu, B.; and Tao, J. 2023. MAE-DFER: Efficient Masked Autoencoder for Self-Supervised Dynamic Facial Expression Recognition . In ACM Int. Conf. Multimedia, 6110--6121

  13. [21]

    Sweldens, W. 1998. The Lifting Scheme : A Construction of Second Generation Wavelets . SIAM Journal on Mathematical Analysis, 29(2): 511--546

  14. [22]

    Tang, H.; and Chai, L. 2024. Facial micro-expression recognition using stochastic graph convolutional network and dual transferred learning. Neural Networks, 178: 106421

  15. [23]

    Tao, Z.; Wang, Y.; Chen, Z.; Wang, B.; Yan, S.; Jiang, K.; Gao, S.; and Zhang, W. 2023. Freq- HD : An Interpretable Frequency-based High-Dynamics Affective Clip Selection Method for in-the- Wild Facial Expression Recognition in Videos . In ACM Int. Conf. Multimedia, 843--852

  16. [24]

    Tran, D.; Bourdev, L.; Fergus, R.; Torresani, L.; and Paluri, M. 2015. Learning Spatiotemporal Features With 3D Convolutional Networks . In Int. Conf. Comput. Vis., 4489--4497

  17. [25]

    Tran, D.; Wang, H.; Torresani, L.; Ray, J.; LeCun, Y.; and Paluri, M. 2018. A Closer Look at Spatiotemporal Convolutions for Action Recognition . In IEEE Conf. Comput. Vis. Pattern Recog., 6450--6459

  18. [26]

    N.; Kaiser, .; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, .; and Polosukhin, I. 2017. Attention is All You Need . Adv. Neural Inform. Process. Syst., 6000--6010

  19. [27]

    Wang, H.; Li, B.; Wu, S.; Shen, S.; Liu, F.; Ding, S.; and Zhou, A. 2023. Rethinking the Learning Paradigm for Dynamic Facial Expression Recognition . In IEEE Conf. Comput. Vis. Pattern Recog., 17958--17968

  20. [28]

    Wang, H.; Mai, X.; Tao, Z.; Tong, X.; Lin, J.; Wang, Y.; Yu, J.; Wang, B.; Yan, S.; Zhao, Q.; Zhou, Z.; Gao, S.; and Zhang, W. 2024. Seeking Certainty In Uncertainty : Dual-Stage Unified Framework Solving Uncertainty in Dynamic Facial Expression Recognition . arXiv:2406.16473

  21. [29]

    Wang, Y.; Sun, Y.; Huang, Y.; Liu, Z.; Gao, S.; Zhang, W.; Ge, W.; and Zhang, W. 2022 a . FERV39k : A Large-Scale Multi-Scene Dataset for Facial Expression Recognition in Videos . In IEEE Conf. Comput. Vis. Pattern Recog., 20890--20899

  22. [30]

    Wang, Y.; Sun, Y.; Song, W.; Gao, S.; Huang, Y.; Chen, Z.; Ge, W.; and Zhang, W. 2022 b . DPCNet : Dual Path Multi-Excitation Collaborative Network for Facial Expression Representation Learning in Videos . In ACM Int. Conf. Multimedia, 101--110

  23. [31]

    Yang, Y.; Jiao, L.; Li, L.; Liu, X.; Liu, F.; Chen, P.; and Yang, S. 2024. LGLFormer : Local -- Global Lifting Transformer for Remote Sensing Scene Parsing . IEEE Trans. Geosci. Remote Sens., 62: 1--13

  24. [32]

    Zhang, Y.; Wang, C.; and Deng, W. 2021. Relative Uncertainty Learning for Facial Expression Recognition . Adv. Neural Inform. Process. Syst., 34: 17616--17627

  25. [33]

    Zhao, Z.; and Liu, Q. 2021. Former- DFER : Dynamic Facial Expression Recognition Transformer . In ACM Int. Conf. Multimedia, 1553--1561

  26. [34]

    Zhao, Z.; Liu, Q.; and Zhou, F. 2021. Robust Lightweight Facial Expression Recognition Network with Label Distribution Training . In AAAI Conf. Artif. Intell., volume 35, 3510--3519

  27. [35]

    Zhao, Z.; and Patras, I. 2023. Prompting Visual-Language Models for Dynamic Facial Expression Recognition . In Brit. Mach. Vis. Conf., 1--14

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.