Pith. sign in

REVIEW 3 major objections 4 minor 41 references

AffectFuse: Cross-Task Feature Fusion with Temporal Modeling for Multi-Task Affective Behavior Analysis

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Post-encoder adaptation with frozen face encoders reaches P=1.7302 on the ABAW multi-task affect recognition benchmark, without training a new face foundation model.

desk verdict An honest, well-ablated ABAW system report; headline P is a validation-selected maximum, so treat the number with care, but it deserves a serious referee. read the letter →

arxiv 2607.16546 v1 pith:3P5YTJPU submitted 2026-07-17 cs.CV

Dipit Saha , Mohammad Raihan Rashid , Shah Mohammad Abdul Mannan , Ahnaf Tahmid , Md. Mehedi Hasan This is my paper
classification cs.CV
keywords multi-tasklearningaffectivebehavioranalysisABAWvalence-arousalfacialactionunitscross-taskfeaturefusionLoRAper-unitexpertrouting
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 tries to establish that for multi-task affective behavior recognition, the encoder is not always the bottleneck: frozen, AffectNet-supervised face encoders can feed a strong multi-task system if the effort is spent after feature extraction. On the ABAW MTL track over s-Aff-Wild2, the proposed pipeline, combining task-specific temporal heads, cross-task feature fusion, a LoRA-adapted action-unit expert, and per-unit expert routing of external data, reaches a validation P of 1.7302 without training a new large-scale face foundation model. The paper argues that valence/arousal and expression gain most from temporal modeling and fusion, while action units repay parameter-efficient fine-tuning and per-unit selection. The result matters because it suggests a cheaper, reproducible route to competitive multi-task affect systems and identifies where frozen transfer saturates.

What carries the argument

The load-bearing mechanism is a staged post-encoder adaptation stack: frozen AffectNet-supervised encoders (EfficientNet-B2 and DDAMFN) provide cached multi-resolution features and remapped expression logits; a self-supervised MAE-Face ViT-B is LoRA-adapted (rank 16, alpha 32, six unfrozen blocks) with a masked asymmetric loss plus a collapse-guarded sigmoid-F1 surrogate to become the AU expert; a 48-frame GRU/BiGRU temporal head and a cross-task fusion module concatenate AU features into valence/arousal and expression representations; and a per-unit expert route (argmax over validation F1 per AU column) selects the best source for each of the twelve action units. Seven-frame smoothing and v

What would settle it

Take the exact validation-selected configuration and run it on the official ABAW test set, or on a held-out split never used in the described ablations: if P drops substantially below 1.7302, or a configuration chosen on a separate development split outperforms it, the central claim that post-encoder adaptation with frozen encoders is sufficient would be weakened.

Watch

Extended reading notes

Core claim

On the official validation split of s-Aff-Wild2, the system reports P=1.7302, composed of average CCC 0.6596 for valence/arousal, macro-F1 0.5097 for eight-class expression, and mean F1 0.5609 across twelve action units, up from P=1.4529 for frozen backbones with plain heads. The central discovery is that after-encoder design choices, not new foundation-model training, drive most of the gain: temporal modeling with a 48-frame window, cross-task fusion that feeds AU features into valence/arousal and expression heads, LoRA fine-tuning of a self-supervised face autoencoder for AU, and per-unit expert routing that lets different AU columns pick different trained sources based on validation F1. T

Load-bearing premise

The load-bearing premise is that choosing every component and fitting the final calibration parameters on the official validation split yields a configuration that transfers to unseen test data; if validation-selected tuning does not generalize, the reported P=1.7302 is an optimistic selected maximum rather than an estimate of true performance.

Editorial extensions

If this is right

  • Competitive multi-task affective systems can be built without training a new face foundation model; frozen AffectNet-supervised features plus task heads, fusion, and temporal modeling account for most of the gains.
  • Action-unit recognition is the part of the pipeline that benefits most from parameter-efficient adaptation and per-unit expert routing; external lab data like DISFA helps as a feature donor but not as direct sequential transfer.
  • Temporal modeling and fusion gains are representation-dependent: increasing the window from 32 to 48 frames after fusion raises the overall score by 0.031.
  • Validation-guided selection with explicit candidate pools and documented negative results, such as flip test-time augmentation hurting all tasks, yields a reproducible ablation path for multi-task affect systems.
  • Unfreezing the encoder for expression and valence/arousal saturates below frozen supervised features, so compute is better spent after the encoder.

Reading between the lines

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

  • Editorial inference: because the same validation split is used both to select every component and to fit the expression bias and AU thresholds, the quoted P=1.7302 is likely an optimistic selected maximum; a held-out test split would quantify how much validation-driven selection overfits.
  • Editorial inference: the per-unit expert-routing idea is not AU-specific; it could be applied to any sparse multi-label task where different labels benefit from different data sources, as long as per-column validation F1 is available.
  • Editorial inference: the flip-TTA negative result hints that horizontal-flip augmentation is harmful for expression and action units because facial asymmetry is informative; a direct test would be to train with horizontal-flip augmentation from the start rather than applying it only at test time.
  • Editorial inference: the frozen-feature cache makes the pipeline data-efficient; the same recipe could be transferred to other in-the-wild face benchmarks by swapping in any supervised face encoder and rerunning the candidate selection.
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 describes AffectFuse, a system for the ABAW 2026 Multi-Task Learning (MTL) track. It builds on frozen AffectNet-supervised EfficientNet-B2 and DDAMFN encoders, a self-supervised MAE-Face encoder adapted via LoRA for action units, temporal GRU/LSTM heads, cross-task feature fusion from AU to valence/arousal and expression, per-unit expert routing with external DISFA data, and validation-selected ensembles. The final configuration achieves P=1.7302 on the official validation split (VA CCC 0.6596, EXPR macro-F1 0.5097, AU mean F1 0.5609). The central claim is that post-encoder adaptation — task heads, fusion, temporal modeling, and per-AU routing — can match or beat retraining large encoders on this benchmark, and that the encoder is not always the bottleneck.

Significance. If the reported validation result transfers to the test set, the contributions are practically significant: a data-efficient pipeline from public frozen checkpoints, a transparent ablation framework, and useful negative results (DISFA sequential transfer, flip-TTA, frozen MAE-Face). The paper is honest about its validation-based selection and includes per-stage diagnostics. However, the quantitative headline and the qualitative 'encoder is not the bottleneck' conclusion rest entirely on a validation-selected maximum with no held-out or test-set confirmation. The significance therefore depends on validation/test agreement, which is not demonstrated in the manuscript.

major comments (3)
  1. [Eqs. (1), (7), (8); Secs. 3.1, 3.7, 3.8] The reported P=1.7302 is obtained by optimizing the same validation metric used for evaluation. Eq. (1) selects each task predictor by argmax over validation M_t; Eq. (7) selects per-AU experts by validation F1; Eq. (8) fits the EXPR class bias b and the AU thresholds θ_j on the validation split after smoothing; Sec. 3.8 selects seeds and top-K ensembles by the same validation scores. With a candidate pool of a dozen-plus families and additional hyperparameters (window length, LoRA rank, unfrozen blocks, etc.), the quoted P is a selected maximum rather than an unbiased estimate of expected performance. The paper's own warning in Sec. 3 ('every added module can overfit validation') is not backed by a nested split, cross-validation, or test-set report. This is load-bearing for both the quantitative claim and the qualitative 'encoder is not the bottleneck' conclusion. Please provide test/le
  2. [Tables 2–11; Sec. 4.3] All reported metrics are point estimates on a single split. Multi-seed runs are used only to form ensembles; no standard deviations or confidence intervals are reported for any row. The final configuration is the argmax over a large validation-driven search, so the gap between validation-selected performance and true test performance is typically nonzero and can be substantial. Without test-set confirmation or an internal holdout, the claim that 'post-encoder adaptation ... provides a strong MTL pipeline' is not fully established. Please add error bars for the main rows and, if possible, a test-set result from the challenge.
  3. [Table 3; Sec. 3.1] The ablation staging is inconsistent. The row labeled 'AU on concat. features + ASL' raises VA from 0.5294 to 0.5870 and EXPR from 0.3948 to 0.4165, even though the stage description indicates only an AU-branch change. The next AU-only row ('AU: LoRA fine-tuned MAE') leaves VA and EXPR unchanged. This suggests the rows do not isolate single candidate families as claimed, or that some joint retraining/feature-cache update is occurring off-stage. Please clarify what is retrained at each stage or correct the table; otherwise the incremental-gain story and the 'compact ablation' contribution are difficult to interpret.
minor comments (4)
  1. [Sec. 3.5; Table 5] The text says 'the complete final four transformer blocks are also optimized,' but Table 5 reports the best result with 6 unfrozen blocks and parameter counts varying from 15.9M to ~44M. Please specify the exact trainable blocks in the final configuration and reconcile the 31.53M-of-89.0M statement with the table.
  2. [Abstract; Sec. 3.7] The code link is a placeholder ('Code is available here') with no URL. Also, 'landmark-aligned224×224 crops' is missing a space.
  3. [Table 2] The 'Organizer score' row is undefined. State what this baseline represents (e.g., challenge baseline or previous best) and cite it.
  4. [Table 1] The 'Pretraining' column entries like 'AffectNet260' and 'AffectNet112' are ambiguous; these appear to be pretraining dataset plus input resolution. Please separate dataset name from resolution.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline P=1.7302 is a validation-selected maximum, not an independent estimate: Eq. (1) and Eq. (8) tune on the same split that is then scored.

  1. fitted input called prediction [Eq. (1) (Sec. 3.1), Eq. (8) (Sec. 3.9), Table 2 (Sec. 4.3)]
    "f ⋆ t = arg max f∈C t Mt(f), P = X t Mt(f ⋆ t) ... The EXPR bias b∈R8 and the 12 AU thresholds θj are fitted on validation after smoothing."

    The reported P is the sum of task metrics M_t evaluated on the official validation split, while Eq. (1) selects each predictor as the argmax of M_t over the candidate pool on that same split, and Eq. (8) fits the EXPR bias and AU thresholds on that same split. Sec. 3.8 additionally selects seed/top-K ensembles by validation scores, and Sec. 3.7 routes per-AU experts by validation F1. Table 2 then quotes P=1.7302 on that split. Thus the headline number is by construction the selected maximum of the validation objective over the explored candidates and calibration parameters, not an unbiased or held-out estimate; the 'prediction' is statistically forced by the selection criterion. The paper itself warns that 'every added module can overfit validation' but provides no nested split or test-set

full rationale

No self-citation chain is load-bearing: references to AffectNet, DDAMFN, MAE-Face, LoRA, DISFA, HSEmotion, and the 7th-edition MTL system are all external, with no overlapping authorship. The feature-extraction and ablation content is therefore not circular in the sense of deriving a result from its own prior work. However, the central quantitative claim is evaluation-circular. Eq. (1) makes the validation metric M_t both the selection criterion (argmax over C_t) and the reported score P; Eq. (8) fits calibration parameters (EXPR bias b, AU thresholds θ_j) on the same validation split; Sec. 3.8 chooses ensembles by validation scores; Sec. 3.7 chooses per-AU experts by validation F1. Reporting P=1.7302 on that same split as evidence that the pipeline is 'strong' turns the headline into a selected maximum rather than a prediction of performance. This is a partial circularity in evaluation, not in the feature derivation, and the paper's explicit caveat about validation overfitting prevents a higher score. Score 6.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard ML metrics, frozen pretrained feature assumptions, validation-based selection, and the usefulness of a lab AU donor. No new physical or theoretical entities are introduced; 'per-unit expert routing' is a selection rule over existing predictors.

free parameters (6)
  • AU thresholds θ_j = 12 values fit on validation (not listed)
    Fitted on the official validation split in Sec. 3.9 to maximize per-AU F1; they are part of final predictions and therefore inflate validation P.
  • EXPR class bias b = 8 values fit on validation (not listed)
    Fitted on validation to shift expression logits; final P measured after this fit.
  • LoRA rank r and alpha = r=16, alpha=32
    Chosen hyperparameters for AU-stage adaptation (Eq. 3); not derived from first principles.
  • Temporal window length L = 48
    Selected by ablation (Table 10); increasing from 32 to 48 improved fused candidates.
  • Ensemble top-K = K=3
    Number of highest-scoring candidates averaged in the final ensembling stage (Sec. 3.8).
  • Sigmoid-F1 guard thresholds = positive rate 0.80; AU F1 0.42
    Hand-set guard rails to prevent the unbounded sigmoid-F1 surrogate from collapsing to all-positive predictions (Sec. 3.5).
assumptions (4)
  • standard math CCC, macro-F1, and multi-label F1 are standard metrics and are accepted as given.
    The paper's objective P (Eq. 1) is an organizer-defined combination of standard metrics; the paper does not need to re-derive them.
  • domain assumption Frozen AffectNet-supervised encoders provide useful feature representations for in-the-wild faces from s-Aff-Wild2.
    Used as the basis of the entire system in Sec. 3.2; empirically supported only by this paper's validation numbers, mostly on one dataset.
  • domain assumption Validation-set model selection via Eq. (1) is a valid way to compose and finalize the MTL system.
    All reported scores are selected on the same validation split; this is the weakest methodological premise and is acknowledged by the authors in Sec. 3.
  • domain assumption DISFA lab data can serve as an AU donor despite the lab-to-wild domain gap.
    Sec. 3.7 shows direct sequential transfer hurts, but donor features are still reused; this donor value is an empirical claim not independently verified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AffectFuse: Cross-Task Feature Fusion with Temporal Modeling for Multi-Task Affective Behavior Analysis." pith.science (2026). https://pith.science/paper/3P5YTJPU

@misc{pith2026260716546,
  author       = {Pith},
  title        = {Pith review of: AffectFuse: Cross-Task Feature Fusion with Temporal Modeling for Multi-Task Affective Behavior Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3P5YTJPU}},
  note         = {Machine review of arXiv:2607.16546}
}
read the original abstract

Affective behavior recognition in the wild requires joint prediction of continuous valence-arousal, categorical facial expression, and multi-label action units from unconstrained face images. We present our system for the Multi-Task Learning (MTL) track of the 11th Affective Behavior Analysis in-the-wild (ABAW) competition on s-Aff-Wild2, the static selected-frame version of Aff-Wild2. The method focuses on post-encoder adaptation: frozen AffectNet-supervised backbones provide multi-resolution features, while task-specific temporal heads and cross-task fusion modules select the useful signals for each target. For action-unit recognition, we adapt MAE-Face with Low-Rank Adaptation (LoRA) and use DISFA through per-unit expert routing rather than direct sequential transfer. Ablations over backbone, temporal, fusion, and AU-adaptation choices define the final configuration. The final system obtains P = 1.7302 on the official validation split, showing that post-encoder adaptation and task-wise modeling choices provide a strong MTL pipeline without training a new large-scale face foundation model.

Figures

Figures reproduced from arXiv: 2607.16546 by the authors.

Figure 1
Figure 1. Overview of AffectFuse. Frozen AffectNet-supervised encoders and an AU-adapted MAE-Face encoder produce per-frame [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. EXPR confusion matrix on validation set. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. AU per-unit F1 on validation set [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 9 linked inside Pith

  1. [1]

    Asymmetric loss for multi-label classification

    Emanuel Ben-Baruch, Tal Ridnik, Nadav Zamir, Asaf Noy, Itamar Friedman, Matan Protter, and Lihi Zelnik-Manor. Asymmetric loss for multi-label classification. InProceed- ings of the IEEE/CVF International Conference on Computer Vision, pages 82–91, 2021. 2

    2021

  2. [2]

    sigmoidf1: A smooth f1 score surrogate loss for multilabel classification.arXiv preprint arXiv:2108.10566,

    Gabriel Bénédict, Vincent Koops, Daan Odijk, and Maarten de Rijke. sigmoidf1: A smooth f1 score surrogate loss for multilabel classification.arXiv preprint arXiv:2108.10566,

    arXiv

  3. [3]

    Learning phrase representations using rnn encoder-decoder for statistical machine translation

    Kyunghyun Cho, Bart van Merriënboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. Learning phrase representations using rnn encoder-decoder for statistical machine translation. InPro- ceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1724–1734,

    2014

  4. [4]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. InInternational Conference on Learning Representa- tions (ICLR), 2021. 2

    2021

  5. [5]

    Facial action coding system.Environmental Psychology & Nonverbal Behavior,

    Paul Ekman and Wallace V Friesen. Facial action coding system.Environmental Psychology & Nonverbal Behavior,

  6. [6]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16000– 16009, 2022. 2

    2022

  7. [7]

    Long short-term memory.Neural Computation, 9(8):1735–1780, 1997

    Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory.Neural Computation, 9(8):1735–1780, 1997. 3

    1997

  8. [8]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. InInterna- tional Conference on Learning Representations (ICLR), 2022. 3

    2022

Show all 41 references
  1. [9]

    Abaw: Valence-arousal estimation, expres- sion recognition, action unit detection & multi-task learning challenges

    Dimitrios Kollias. Abaw: Valence-arousal estimation, expres- sion recognition, action unit detection & multi-task learning challenges. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2328–2336,

  2. [10]

    Abaw: Learning from synthetic data & multi-task learning challenges

    Dimitrios Kollias. Abaw: Learning from synthetic data & multi-task learning challenges. InEuropean Conference on Computer Vision, pages 157–172. Springer, 2023. 2

    2023

  3. [11]

    Aff-wild2: Extend- ing the aff-wild database for affect recognition.arXiv preprint arXiv:1811.07770, 2018

    Dimitrios Kollias and Stefanos Zafeiriou. Aff-wild2: Extend- ing the aff-wild database for affect recognition.arXiv preprint arXiv:1811.07770, 2018. 4

    2018 arXiv

  4. [12]

    Expression, affect, action unit recognition: Aff-wild2, multi-task learning and arcface.arXiv preprint arXiv:1910.04855, 2019

    Dimitrios Kollias and Stefanos Zafeiriou. Expression, affect, action unit recognition: Aff-wild2, multi-task learning and arcface.arXiv preprint arXiv:1910.04855, 2019. 1, 2

    1910 arXiv

  5. [13]

    Affect analysis in-the-wild: Valence-arousal, expressions, action units and a unified framework.arXiv preprint arXiv:2103.15792, 2021

    Dimitrios Kollias and Stefanos Zafeiriou. Affect analysis in-the-wild: Valence-arousal, expressions, action units and a unified framework.arXiv preprint arXiv:2103.15792, 2021. 2

    2021 arXiv

  6. [14]

    Analysing affective behavior in the second abaw2 competition

    Dimitrios Kollias and Stefanos Zafeiriou. Analysing affective behavior in the second abaw2 competition. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 3652–3660, 2021. 2

    2021

  7. [15]

    Face behavior a la carte: Expressions, af- fect and action units in a single network.arXiv preprint arXiv:1910.11111, 2019

    Dimitrios Kollias, Viktoriia Sharmanska, and Stefanos Zafeiriou. Face behavior a la carte: Expressions, af- fect and action units in a single network.arXiv preprint arXiv:1910.11111, 2019. 2

    1910 arXiv

  8. [16]

    Deep affect prediction in-the-wild: Aff-wild database and challenge, deep architec- tures, and beyond.International Journal of Computer Vision, pages 1–23, 2019

    Dimitrios Kollias, Panagiotis Tzirakis, Mihalis A Nicolaou, Athanasios Papaioannou, Guoying Zhao, Björn Schuller, Irene Kotsia, and Stefanos Zafeiriou. Deep affect prediction in-the-wild: Aff-wild database and challenge, deep architec- tures, and beyond.International Journal o...

    2019

  9. [17]

    Analysing affective behavior in the first abaw 2020 competition

    Dimitrios Kollias, Attila Schulc, Elnar Hajiyev, and Stefanos Zafeiriou. Analysing affective behavior in the first abaw 2020 competition. In2020 15th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2020), pages 794–800. IEEE, 2020. 2

    2020

  10. [18]

    Distribution matching for heterogeneous multi- task learning: a large-scale face study.arXiv preprint arXiv:2105.03790, 2021

    Dimitrios Kollias, Viktoriia Sharmanska, and Stefanos Zafeiriou. Distribution matching for heterogeneous multi- task learning: a large-scale face study.arXiv preprint arXiv:2105.03790, 2021. 2

    2021 arXiv

  11. [19]

    Abaw: Valence-arousal esti- mation, expression recognition, action unit detection & emo- tional reaction intensity estimation challenges

    Dimitrios Kollias, Panagiotis Tzirakis, Alice Baird, Alan Cowen, and Stefanos Zafeiriou. Abaw: Valence-arousal esti- mation, expression recognition, action unit detection & emo- tional reaction intensity estimation challenges. InProceedings of the IEEE/CVF Conference on Comput...

    2023

  12. [20]

    Behaviour4all: in-the-wild facial behaviour analysis toolkit.arXiv preprint arXiv:2409.17717, 2024

    Dimitrios Kollias, Chunchang Shao, Odysseus Kaloidas, and Ioannis Patras. Behaviour4all: in-the-wild facial behaviour analysis toolkit.arXiv preprint arXiv:2409.17717, 2024. 2

    2024 arXiv

  13. [21]

    Distribution matching for multi-task learning of classification tasks: a large-scale study on faces & beyond

    Dimitrios Kollias, Viktoriia Sharmanska, and Stefanos Zafeiriou. Distribution matching for multi-task learning of classification tasks: a large-scale study on faces & beyond. In Proceedings of the AAAI Conference on Artificial Intelligence, pages 2813–2821, 2024. 2

    2024

  14. [22]

    The 6th affective behavior analysis in-the-wild (abaw) competition

    Dimitrios Kollias, Panagiotis Tzirakis, Alan Cowen, Stefanos Zafeiriou, Irene Kotsia, Alice Baird, Chris Gagne, Chun- chang Shao, and Guanyu Hu. The 6th affective behavior analysis in-the-wild (abaw) competition. InProceedings of the IEEE/CVF Conference on Computer Vision and ...

    2024

  15. [23]

    7th abaw competition: Multi-task learning and compound expression recognition

    Dimitrios Kollias, Stefanos Zafeiriou, Irene Kotsia, Abhinav Dhall, Shreya Ghosh, Chunchang Shao, and Guanyu Hu. 7th abaw competition: Multi-task learning and compound expression recognition. InEuropean Conference on Computer Vision, pages 31–45. Springer, 2024

    2024

  16. [24]

    Advancements in affective and behavior analysis: The 8th abaw workshop and competition

    Dimitrios Kollias, Panagiotis Tzirakis, Alan Cowen, Stefanos Zafeiriou, Irene Kotsia, Eric Granger, Marco Pedersoli, Simon Bacon, Alice Baird, Chris Gagne, et al. Advancements in affective and behavior analysis: The 8th abaw workshop and competition. InProceedings of the Compu...

    2025

  17. [25]

    From emotions to violence: Multimodal fine-grained behavior analysis at the 9th abaw

    Dimitrios Kollias, Stefanos Zafeiriou, Irene Kotsia, Greg Slabaugh, Damith Chamalke Senadeera, Jianian Zheng, Kaushal Kumar Keshlal Yadav, Chunchang Shao, and Guanyu Hu. From emotions to violence: Multimodal fine-grained behavior analysis at the 9th abaw. InProceedings of the ...

    2025

  18. [26]

    From affect to complex behavior: Advancing multimodal human-centered ai at the 10th abaw workshop & competition, 2026

    Dimitrios Kollias, Panagiotis Tzirakis, Alan Cowen, Stefanos Zafeiriou, Irene Kotsia, Eric Granger, Marco Pedersoli, Simon Bacon, Jens Madsen, Soufiane Belharbi, Muhammad Haseeb Aslam, Chunchang Shao, and Guanyu Hu. From affect to complex behavior: Advancing multimodal human-c...

    2026

  19. [27]

    A concordance correlation coefficient to evaluate reproducibility.Biometrics, 45(1):255–268, 1989

    Lawrence I-Kuei Lin. A concordance correlation coefficient to evaluate reproducibility.Biometrics, 45(1):255–268, 1989. 2

    1989

  20. [28]

    Affective behaviour analysis via progressive learning.arXiv preprint arXiv:2407.16945, 2024

    Chen Liu, Wei Zhang, Feng Qiu, Lincheng Li, Dadong Wang, and Xin Yu. Affective behaviour analysis via progressive learning.arXiv preprint arXiv:2407.16945, 2024. 2, 3

    2024 arXiv

  21. [29]

    Poster++: A simpler and stronger facial expression recognition network.Pattern Recognition, 157:110951, 2025

    Jiawei Mao, Rui Xu, Xuesong Yin, Yuanqi Chang, Binling Nie, and Aibin Huang. Poster++: A simpler and stronger facial expression recognition network.Pattern Recognition, 157:110951, 2025. 5

    2025

  22. [30]

    Mohammad Mavadati, Mohammad H

    S. Mohammad Mavadati, Mohammad H. Mahoor, Kevin Bartlett, and Philip Trinh. Automatic detection of non-posed facial action units. In2012 19th IEEE International Confer- ence on Image Processing (ICIP), pages 1817–1820. IEEE,

  23. [31]

    Mohammad Mavadati, Mohammad H

    S. Mohammad Mavadati, Mohammad H. Mahoor, Kevin Bartlett, Philip Trinh, and Jeffrey F. Cohn. Disfa: A sponta- neous facial action intensity database.IEEE Transactions on Affective Computing, 4(2):151–160, 2013. 2, 4

    2013

  24. [32]

    Ma- hoor

    Ali Mollahosseini, Behzad Hasani, and Mohammad H. Ma- hoor. Affectnet: A database for facial expression, valence, and arousal computing in the wild.IEEE Transactions on Affective Computing, 10(1):18–31, 2017. 2

    2017

  25. [33]

    Hosseini, Timo- thy D

    Arsha Pourramezan Fard, Mohammad M. Hosseini, Timo- thy D. Sweeny, and Mohammad H. Mahoor. Affectnet+: A database for enhancing facial expression recognition with soft-labels.IEEE Transactions on Affective Computing, 2025. 4

    2025

  26. [34]

    A circumplex model of affect.Journal of Personality and Social Psychology, 39(6):1161–1178, 1980

    James A Russell. A circumplex model of affect.Journal of Personality and Social Psychology, 39(6):1161–1178, 1980. 1

    1980

  27. [35]

    Savchenko

    Andrey V . Savchenko. Video-based frame-level facial analy- sis of affective behavior on mobile devices using efficientnets. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 2359–2366,

  28. [36]

    Savchenko

    Andrey V . Savchenko. Mt-emotieffnet for multi-task human affective behavior analysis and learning from synthetic data. InEuropean Conference on Computer Vision Workshops (EC- CVW), pages 45–59. Springer, 2023. 2

    2023

  29. [37]

    Savchenko

    Andrey V . Savchenko. Hsemotion team at the 7th abaw chal- lenge: Multi-task learning and compound facial expression recognition.arXiv preprint arXiv:2407.13184, 2024. 2

    2024 arXiv

  30. [38]

    Efficientnet: Rethinking model scaling for convolutional neural networks

    Mingxing Tan and Quoc Le. Efficientnet: Rethinking model scaling for convolutional neural networks. InInternational Conference on Machine Learning (ICML), pages 6105–6114. PMLR, 2019. 2

    2019

  31. [39]

    Aff-wild: Valence and arousal in-the-wild challenge

    Stefanos Zafeiriou, Dimitrios Kollias, Mihalis A Nicolaou, Athanasios Papaioannou, Guoying Zhao, and Irene Kotsia. Aff-wild: Valence and arousal in-the-wild challenge. In2017 IEEE Conference on Computer Vision and Pattern Recogni- tion Workshops (CVPRW), pages 1980–1987. IEEE, 2017. 1

    1980

  32. [40]

    A dual-direction attention mixed feature net- work for facial expression recognition.Electronics, 12(17): 3595, 2023

    Saining Zhang, Yuhang Zhang, Ye Zhang, Yufei Wang, and Zhigang Song. A dual-direction attention mixed feature net- work for facial expression recognition.Electronics, 12(17): 3595, 2023. 2

    2023

  33. [41]

    Multi- modal facial affective analysis based on masked autoencoder

    Wei Zhang, Bowen Ma, Feng Qiu, and Yu Ding. Multi- modal facial affective analysis based on masked autoencoder. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5792–5801, 2023. 2

    2023

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.