Pith. sign in

REVIEW 4 major objections 6 minor 33 references

Multi-level and Multi-modal Action Anticipation

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Self-generated fine-grained text lifts action anticipation by 3.08%.

desk verdict The paper omits the main training loss and its ablations contradict the headline 3.08% gain; the underlying idea is plausible enough to warrant peer review with clear demands for revision. read the letter →

arxiv 2506.02382 v1 pith:XY3EKU7Y submitted 2025-06-03 cs.CV cs.LG

Seulgi Kim , Ghazal Kaviani , Mohit Prabhushankar , Ghassan AlRegib This is my paper · ORCID
classification cs.CVcs.LG
keywords actionanticipationmulti-modallearningfine-grainedtextgenerationtemporalconsistencylosshierarchicalsemanticmodelinglong-termvideounderstanding
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 sets out to show that action anticipation from partially observed video improves when the model is given a second, textual modality alongside the visual stream, and that this text does not have to come from an external source. The proposed m&m-Ant method generates fine-grained textual labels from the video features themselves, using a temporal consistency loss to separate repeated occurrences of the same coarse action into distinct sub-action clusters. These fine-grained texts are concatenated with video embeddings and then aligned with coarse segmentation labels through cross-attention, forming a hierarchical multi-modal representation. On Breakfast, 50 Salads, and DARai, the authors report state-of-the-art accuracy with an average gain of 3.08% over prior methods, with the largest improvements on dense datasets and at low observation rates. A sympathetic reader would take the paper as establishing that a self-produced textual modality plus hierarchical semantic modeling is a practical and effective route to better anticipation without external text resources.

What carries the argument

The load-bearing mechanism is the fine-grained text generator and its temporal consistency loss. Frames from the same action class that lie within a continuous temporal interval are grouped into a cluster, and the loss $L_{\text{tcl}} = \lambda_1 L_{\text{intra}} + \lambda_2 L_{\text{inter}}$ is applied, where $L_{\text{intra}} = \sum_k \sum_{x_i \in X^{(k)}} \lVert x_i - \mu_k \rVert^2$ enforces cohesion toward the cluster centroid $\mu_k$ and $L_{\text{inter}} = \sum_{k, k' \neq k} 1/\lVert \mu_k - \mu_{k'} \rVert$ penalizes clusters that are close, so repeated instances of the same nominal action separate into distinct clusters. The cluster assignments are rendered as fine-grained text tokens, concatenated with video features, and aligned with coarse segmentation labels through multi-head cross-attention in the anticipation module. This chain from raw features to clustered text to cross-modal attention is what carries the claimed accuracy improvement.

What would settle it

Train m&m-Ant with the fine-grained text generator replaced by a fixed, untrained random projection of the same video features (or by the coarse labels copied verbatim) while keeping all other components and parameter counts identical; if the 3.08% average improvement over FUTR persists, the reported gain is not caused by the learned fine-grained textual semantics. Alternatively, measure the conditional mutual information between generated text and future action labels given observed video features; near-zero values would indicate the text carries no anticipatory information beyond the video.

Watch

Extended reading notes

Core claim

The central claim is that a video-only anticipation system can be turned into a genuinely multi-modal system by generating fine-grained textual labels from the video's own features, and that these labels carry predictive signal beyond both raw visual features and coarse segmentation labels. Concretely, m&m-Ant concatenates sampled video embeddings with the generated fine-grained text, processes the concatenation through multi-head self-attention, and then injects coarse segmentation labels as keys and values in a multi-head cross-attention layer to encode high-level semantic structure. The fine-grained text generator is trained with a cross-entropy term plus a temporal consistency loss: an intra-cluster term pulls frames of the same action inside a continuous temporal interval toward their cluster centroid, while an inter-cluster term pushes apart clusters that are temporally distant even when they share the same class label. The paper argues that this recovers sub-action flows that frame-level classification ignores, and presents experiments on three datasets as evidence that the recovered signal is enough to beat the FUTR, GTD, and AFFT baselines. If the claim is right, the implication is that the bottleneck in long-term action anticipation is not the absence of external linguistic knowledge but the model's ability to organize observed video into semantically meaningful temporal units.

Load-bearing premise

The method assumes that the fine-grained text it generates from video features is a genuinely new semantic signal, not just a re-encoding of information the video features and coarse labels already contain; if that assumption fails, the multi-modal architecture gains nothing over a single-modality model with more parameters.

Editorial extensions

If this is right

  • At low observation rates (for example $\alpha = 0.1$), where visual input is scarce, the fine-grained text lets the model exploit subtle cues that a visual-only model misses, yielding larger relative gains.
  • On dense, transition-heavy datasets such as 50 Salads, the improvement over the FUTR baseline is larger than on sparser datasets, because the fine-grained text captures sub-action nuances and frequent transitions.
  • Ablations in the paper show that the multi-modal variant beats a uni-modal variant at every observation rate, and that adding fine-grained text on top of coarse labels further improves accuracy, especially when little of the video has been observed.
  • On the untrimmed, hierarchically labeled DARai dataset, the method stays ahead of FUTR at every reported $\alpha$ and $\beta$ setting, with a maximum reported gain of 2.27%.

Reading between the lines

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

  • Because the 'text' modality is generated from the same video features, part of the gain may come from added model capacity or a regularizing auxiliary task; a matched-capacity single-modality baseline with the same parameter count would separate those effects.
  • The temporal consistency loss is a general recipe for sequence labeling: any task where the same nominal class recurs at widely separated times could benefit from the same intra-cluster/inter-cluster separation objective.
  • A natural next experiment is to add a genuine external text source, such as recipe instructions or narrated transcripts; if accuracy rises further, it would confirm that semantic text, not just the generator, is driving the improvement.
  • The inter-cluster term uses an inverse distance $1/\lVert \mu_k - \mu_{k'} \rVert$, which diverges as clusters coincide; a bounded or capped repulsion term would make training more stable and would test how much of the reported gain depends on this particular formulation.
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

4 major / 6 minor

Summary. The paper proposes m&m-Ant, a multi-modal and multi-level framework for long-term action anticipation. It combines visual features with a generated fine-grained text modality and coarse segmentation labels, using a video encoder, a video segmentation module, a fine-grained text generator trained with a temporal consistency loss, and a multi-modal attention module. The claimed contribution is a 3.08% average accuracy improvement over state-of-the-art on Breakfast, 50 Salads, and DARai. The paper includes ablation studies on the effect of the multi-modal fusion and the multi-level (hierarchical) modeling.

Significance. If the results hold, the idea of generating fine-grained textual semantics from visual features and fusing them with video and coarse labels in a cross-attention architecture is a plausible and potentially useful direction for long-term action anticipation. The paper provides a public code link and evaluates on three datasets, including a recently introduced DARai dataset. However, the manuscript as written has major reproducibility gaps (no anticipation loss is specified) and internal inconsistencies in the ablation tables that directly contradict the stated claims, so the empirical evidence cannot currently be accepted at face value.

major comments (4)
  1. [Sections 3.2-3.4 and 4.1] No training objective is defined for the action anticipation module or the video segmentation module. Equation (3) defines a loss only for the fine-grained text generator (L_total = L_ce + L_tcl). Section 3.4 describes the forward computations of the multi-modal action anticipation module (H_input, MHSA, MHCS) but gives no loss over future action labels, and Section 4.1 lists only optimizer settings (AdamW, 60 epochs, learning rate 1e-3). Without an anticipation loss, the model cannot be trained as described, and the reported accuracy gains cannot be reproduced or attributed to the proposed architecture.
  2. [Tables 3 and 4 and Section 4.2] The text claims that the multi-modal approach 'consistently outperforms the uni-modal counterpart across all observation rates,' but Table 3 shows the opposite on 50 Salads at observation rate 0.3: uni-modal 90.92 vs. multi-modal 84.55, a reversal of 6.37 points. Table 4 similarly shows w/o multi-level 87.91 vs. w/ multi-level 84.55 at the same setting. The non-monotonic pattern in the uni-modal column (90.92 at 0.3 dropping to 73.92 at 0.4) also suggests a possible mislabeling or evaluation error. These contradictions undermine the central consistency claim and must be resolved, either by correcting the tables/text or by rerunning the experiments.
  3. [Section 3.2 and Figure 2] The textual modality is generated entirely from the same video features and dataset labels, with no external text source, recipe input, or language model. The fine-grained text generator first clusters video features and then classifies them into fine-grained labels. The claimed multi-modal complementarity is therefore not independently grounded: the 'text' may be just a re-encoding of the visual features. The paper should include a control experiment that replaces the fine-grained text generator with an equally expressive visual feature transformation (e.g., an MLP on the same video features) or replaces the text embeddings with one-hot fine-grained labels, to show that the improvement is due to cross-modal fusion rather than added parameters or label conditioning.
  4. [Section 4.1 and Tables 1-4] The paper states that results are averaged over 3 runs with fixed seeds 1, 10, 13452, but no standard deviations or confidence intervals are reported in any table. Several reported differences are small (e.g., DARai, alpha=0.2, beta=0.1: FUTR 25.05 vs. m&m-Ant 25.75), so without error bars the central claim of a 3.08% average improvement cannot be assessed statistically. Reporting variance or per-seed results is necessary to establish that the gains are not noise.
minor comments (6)
  1. [Section 3.2, Eq. (2)] The inter-cluster separation loss L_inter = sum_{k,k' != k} 1/||mu_k - mu_k'|| is unbounded as cluster centroids approach each other and its scale depends on the number of clusters; the paper does not discuss the choice of lambda_1 and lambda_2 or their sensitivity.
  2. [Section 3.2] The clustering procedure is underspecified: the paper says frames of the same class within a continuous temporal interval are grouped into a cluster, but it does not state how intervals are determined (e.g., connected components), how many clusters are used, or what happens when the same class appears in disjoint intervals.
  3. [Table 2] The entry '25.7024.24' lacks a separating space and should read '25.70 24.24'; the same formatting problem appears in the row labels of Tables 3 and 4 (e.g., 'Multi-modal30.05').
  4. [Section 4.1] The phrase 'average performance across 3 number of iteration' is a typo; it should read '3 iterations' or '3 runs'.
  5. [Section 4.2 and Table 1] The paper does not specify how the reported 3.08% average improvement is computed across datasets and evaluation settings; the aggregation rule should be stated explicitly.
  6. [Table 1] On 50 Salads at alpha=0.3, beta=0.1, both FUTR and m&m-Ant are reported at 100.00 accuracy, which is suspicious and suggests a saturated or degenerate evaluation setting; this should be discussed.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: the 'textual' modality is generated from the same video features and dataset labels, so the claimed multi-modal gain is self-referential; the held-out future-action evaluation keeps the SOTA result from being fully circular.

  1. self definitional [Section 2.2; Eq. (3); Section 3.4]
    "we introduce a fine-grained text generator that first extracts fine-grained visual features and then translates them into semantically rich texts to help disambiguate uncertainty in action anticipation. ... Hinput = Concat(Hvideo, Hfine-grained)."

    The fine-grained text Hfine-grained is not an external textual modality; it is the output of a generator trained on the same video features Fτ (Eq. 1) using Ltotal = Lce + Ltcl (Eq. 3), where Lce is cross-entropy on the dataset's fine-grained labels. At fusion, the model uses Hinput = Concat(Hvideo, Hfine-grained), so the 'textual' stream is a learned function of the visual stream and the label vocabulary. Hence, the paper's claim that integrating textual cues drives the gains reduces by construction to 'visual features plus an embedding of predicted fine-grained labels of the same visual features.' The complementarity of the two modalities is therefore self-defined rather than independently sourced.

full rationale

m&m-Ant's final accuracy numbers are computed against held-out future action labels, so the headline SOTA claim is not circular by construction. The main circularity concern is the 'textual modality': the fine-grained text is generated from the same video features and the dataset's fine-grained labels (Eqs. 1 and 3), then concatenated with video features (Section 3.4). Consequently, the claimed multi-modal complementarity reduces to a self-generated label embedding of the visual stream; there is no external text source, and the 'multi-modal' framing renames a within-model prediction as a second modality. This weakens the interpretation of the 3.08% gain but does not make the future-anticipation evaluation itself circular. Separately, the paper omits the training objective for the video segmentation and anticipation modules (only Ltotal for the text generator is given), and the ablations in Tables 3 and 4 contain entries (50 Salads, α=0.3) that contradict the claimed consistent gains; these are consistency/reproducibility defects, not circularity. Self-citations [14] and [24] are present but not load-bearing for the core derivation.

Assumptions & free parameters 4 free parameters · 4 assumptions · 1 invented entities

The central claim rests on several unstated domain assumptions about label availability and feature sufficiency, plus unreported hyperparameters that control the proposed loss. The only invented entity is the internally generated 'text' modality, which lacks independent grounding.

free parameters (4)
  • lambda1, lambda2 (temporal consistency loss weights) = Not reported
    Equation (2) weights intra-cluster and inter-cluster terms; the paper gives no values, ranges, or sensitivity analysis. These directly control the clustering behavior that produces the 'fine-grained text' used by the anticipation module.
  • Temporal stride tau = 3 (Breakfast), 6 (50 Salads), 15 (DARai)
    Chosen per dataset to align temporal resolution; no ablation or criterion is given.
  • Hidden dimension D = 128 (Breakfast/DARai), 512 (50 Salads)
    Architecture capacity is chosen per dataset with no ablation.
  • Number of queries = 8
    Set without justification in Section 4.1; no sensitivity study is reported.
assumptions (4)
  • domain assumption Frame-level action labels are available for training the segmentation module, the fine-grained text generator, and the anticipation module.
    Section 3.2 uses ground-truth class labels to form clusters; Section 3.3 trains a segmentation head. The method assumes this supervision exists for all training videos.
  • domain assumption The dataset's fine-grained labels are semantically reliable enough to serve as 'textual' guidance.
    The fine-grained text generator is trained with cross-entropy on fine-grained labels (Section 3.2), and these outputs are treated as complementary semantic text.
  • domain assumption Pre-extracted ResNet features are a sufficient visual representation for both segmentation and anticipation.
    Section 4.1 states pre-extracted ResNet features are used as input; no end-to-end visual learning is attempted.
  • domain assumption Mean-over-classes (MoC) accuracy is the appropriate protocol metric.
    Section 4.1 follows the protocol of [7], which is standard for long-term anticipation.
invented entities (1)
  • Fine-grained text (generated textual modality)
    purpose: Provides fine-grained semantic guidance to cross-attention in the anticipation module (Section 3.4).
    The 'text' is generated by clustering video features and classifying them into dataset labels; no external text corpus, language model, or independent textual input is used. The multi-modal claim therefore rests on an internal signal.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-level and Multi-modal Action Anticipation." pith.science (2026). https://pith.science/paper/XY3EKU7Y

@misc{pith2026250602382,
  author       = {Pith},
  title        = {Pith review of: Multi-level and Multi-modal Action Anticipation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XY3EKU7Y}},
  note         = {Machine review of arXiv:2506.02382}
}
read the original abstract

Action anticipation, the task of predicting future actions from partially observed videos, is crucial for advancing intelligent systems. Unlike action recognition, which operates on fully observed videos, action anticipation must handle incomplete information. Hence, it requires temporal reasoning, and inherent uncertainty handling. While recent advances have been made, traditional methods often focus solely on visual modalities, neglecting the potential of integrating multiple sources of information. Drawing inspiration from human behavior, we introduce \textit{Multi-level and Multi-modal Action Anticipation (m\&m-Ant)}, a novel multi-modal action anticipation approach that combines both visual and textual cues, while explicitly modeling hierarchical semantic information for more accurate predictions. To address the challenge of inaccurate coarse action labels, we propose a fine-grained label generator paired with a specialized temporal consistency loss function to optimize performance. Extensive experiments on widely used datasets, including Breakfast, 50 Salads, and DARai, demonstrate the effectiveness of our approach, achieving state-of-the-art results with an average anticipation accuracy improvement of 3.08\% over existing methods. This work underscores the potential of multi-modal and hierarchical modeling in advancing action anticipation and establishes a new benchmark for future research in the field. Our code is available at: https://github.com/olivesgatech/mM-ant.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 30 canonical work pages

  1. [1]

    checking the refrigerator

    INTRODUCTION Action anticipation is a critical task for advancing intelli- gent systems, where an algorithm processes a partially ob- served video to predict a sequence of future actions by uti- lizing the available visual context [1, 2, 3]. In contrast to action recognition [4, 5, 6], which focuses on identifying ac- tions from fully observed videos, act...

    work page

  2. [2]

    Long-term Action Anticipation Action anticipation emphasizes predicting potential upcom- ing actions before they are observed

    RELATED WORK 2.1. Long-term Action Anticipation Action anticipation emphasizes predicting potential upcom- ing actions before they are observed. The availability of large-scale video datasets [14, 15, 16] has spurred significant progress in addressing the challenge of action anticipation. Action anticipation methods can be broadly categorized into Fig. 2 ...

    work page

  3. [3]

    We be- gin with the Video Encoder, which extracts meaningful fea- tures from the input video

    METHODOLOGY This section details the components of m&m-Ant. We be- gin with the Video Encoder, which extracts meaningful fea- tures from the input video. The Video Segmentation Mod- ule produces initial segmentation labels for each input frame. To address the limitations and uncertainties of these segmen- tation labels, we introduce the Fine-grained Text ...

    work page

  4. [4]

    EXPERIMENTS AND ANALYSIS In this section, we provide an overview of the datasets and experimental setups employed in our experiments, followed by detailed analyses of the proposed method. 4.1. Datasets and Implementation Details Datasets. We aim to evaluate our model’s ability to handle diverse environments, sparse temporal patterns, and dense ac- tion se...

    work page arXiv

  5. [5]

    By incorporating a fine-grained text generator with a novel loss function, our method ad- dresses the challenges of inaccurate video labels, leading to improved prediction accuracy

    CONCLUSION In conclusion, we introduce m&m-Ant, a novel multi-modal action anticipation approach that effectively integrates both visual and textual information to predict future actions from partially observed videos. By incorporating a fine-grained text generator with a novel loss function, our method ad- dresses the challenges of inaccurate video label...

    work page

  6. [6]

    ActFusion: a Unified Diffusion Model for Action Segmentation and Anticipation

    Dayoung Gong, Suha Kwak, and Minsu Cho, “Actfu- sion: a unified diffusion model for action segmentation and anticipation,” arXiv:2412.04353, 2024

    work page Pith review arXiv 2024

  7. [7]

    Future transformer for long-term action anticipation,

    Dayoung Gong, Joonseok Lee, Manjin Kim, Seong Jong Ha, and Minsu Cho, “Future transformer for long-term action anticipation,” in CVPR, 2022, pp. 3052–3061

    work page 2022

  8. [8]

    Object-centric video representation for long-term ac- tion anticipation,

    Ce Zhang, Changcheng Fu, Shijie Wang, Nakul Agar- wal, Kwonjoon Lee, Chiho Choi, and Chen Sun, “Object-centric video representation for long-term ac- tion anticipation,” in WACV, 2024, pp. 6751–6761

    work page 2024

Show all 33 references
  1. [9]

    Hierarchical compositional represen- tations for few-shot action recognition,

    Changzhen Li, Jie Zhang, Shuzhe Wu, Xin Jin, and Shiguang Shan, “Hierarchical compositional represen- tations for few-shot action recognition,” CVIU, 2024

    2024

  2. [10]

    Uncertainty-based spatial-temporal attention for online action detection,

    Hongji Guo, Zhou Ren, Yi Wu, Gang Hua, and Qiang Ji, “Uncertainty-based spatial-temporal attention for online action detection,” in ECCV. Springer, 2022, pp. 69–86

    2022

  3. [11]

    Soar: Scene-debiasing open-set action recognition,

    Yuanhao Zhai, Ziyi Liu, Zhenyu Wu, Yi Wu, Chunluan Zhou, David Doermann, Junsong Yuan, and Gang Hua, “Soar: Scene-debiasing open-set action recognition,” in ICCV, 2023, pp. 10244–10254

    2023

  4. [12]

    When will you do what?-anticipating temporal occur- rences of activities,

    Yazan Abu Farha, Alexander Richard, and Juergen Gall, “When will you do what?-anticipating temporal occur- rences of activities,” in CVPR, 2018, pp. 5343–5352

    2018

  5. [13]

    Time- conditioned action anticipation in one shot,

    Qiuhong Ke, Mario Fritz, and Bernt Schiele, “Time- conditioned action anticipation in one shot,” in CVPR, 2019, pp. 9925–9934

    2019

  6. [14]

    Temporal aggregate representations for long-range video understanding,

    Fadime Sener, Dipika Singhania, and Angela Yao, “Temporal aggregate representations for long-range video understanding,” in ECCV. Springer, 2020

    2020

  7. [15]

    Long-term anticipation of activities with cy- cle consistency,

    Yazan Abu Farha, Qiuhong Ke, Bernt Schiele, and Juer- gen Gall, “Long-term anticipation of activities with cy- cle consistency,” in PR. Springer, 2021, pp. 159–173

    2021

  8. [16]

    Uncertainty-aware action decou- pling transformer for action anticipation,

    Hongji Guo, Nakul Agarwal, Shao-Yuan Lo, Kwonjoon Lee, and Qiang Ji, “Uncertainty-aware action decou- pling transformer for action anticipation,” in CVPR, 2024, pp. 18644–18654

    2024

  9. [17]

    The lan- guage of actions: Recovering the syntax and semantics of goal-directed human activities,

    Hilde Kuehne, Ali Arslan, and Thomas Serre, “The lan- guage of actions: Recovering the syntax and semantics of goal-directed human activities,” in CVPR, 2014

    2014

  10. [18]

    Combining embedded accelerometers with computer vision for rec- ognizing food preparation activities,

    Sebastian Stein and Stephen J McKenna, “Combining embedded accelerometers with computer vision for rec- ognizing food preparation activities,” in Proceedings of the 2013 ACM international joint conference on Perva- sive and ubiquitous computing , 2013, pp. 729–738

    2013

  11. [19]

    Hierarchical and multimodal data for daily activity understanding,

    Ghazal Kaviani, Yavuz Yarici, Seulgi Kim, Mohit Prabhushankar, Ghassan AlRegib, Mashhour Solh, and Ameya Patil, “Hierarchical and multimodal data for daily activity understanding,” arXiv preprint arXiv:2504.17696, 2025

    2025 arXiv

  12. [20]

    Rescaling egocentric vision: Col- lection, pipeline and challenges for epic-kitchens-100,

    Dima Damen, Hazel Doughty, Giovanni Maria Farinella, Antonino Furnari, Evangelos Kazakos, Jian Ma, Davide Moltisanti, Jonathan Munro, Toby Perrett, Will Price, et al., “Rescaling egocentric vision: Col- lection, pipeline and challenges for epic-kitchens-100,” IJCV, pp. 1–23, 2022

    2022

  13. [21]

    Ego4d: Around the world in 3,000 hours of egocentric video,

    Kristen Grauman, Andrew Westbury, Eugene Byrne, Zachary Chavis, Antonino Furnari, Rohit Girdhar, Jack- son Hamburger, Hao Jiang, Miao Liu, Xingyu Liu, et al., “Ego4d: Around the world in 3,000 hours of egocentric video,” in CVPR, 2022, pp. 18995–19012

    2022

  14. [22]

    Anticipating hu- man actions by correlating past with the future with jac- card similarity measures,

    Basura Fernando and Samitha Herath, “Anticipating hu- man actions by correlating past with the future with jac- card similarity measures,” in CVPR, 2021, pp. 13224– 13233

    2021

  15. [23]

    Uncertainty-aware anticipation of activities,

    Yazan Abu Farha and Juergen Gall, “Uncertainty-aware anticipation of activities,” in ICCVW, 2019, pp. 0–0

    2019

  16. [24]

    Weakly supervised action learning with rnn based fine- to-coarse modeling,

    Alexander Richard, Hilde Kuehne, and Juergen Gall, “Weakly supervised action learning with rnn based fine- to-coarse modeling,” in CVPR, 2017, pp. 754–763

    2017

  17. [25]

    Ms-tcn: Multi-stage temporal convolutional network for action segmenta- tion,

    Yazan Abu Farha and Jurgen Gall, “Ms-tcn: Multi-stage temporal convolutional network for action segmenta- tion,” in CVPR, 2019, pp. 3575–3584

    2019

  18. [26]

    Leveraging uncertainty to rethink loss functions and evaluation measures for egocentric action anticipation,

    Antonino Furnari, Sebastiano Battiato, and Giovanni Maria Farinella, “Leveraging uncertainty to rethink loss functions and evaluation measures for egocentric action anticipation,” in ECCVW, 2018, pp. 0–0

    2018

  19. [27]

    Predictive uncer- tainty estimation via prior networks,

    Andrey Malinin and Mark Gales, “Predictive uncer- tainty estimation via prior networks,” Advances in neu- ral information processing systems, vol. 31, 2018

    2018

  20. [28]

    Learning the predictability of the future,

    D ´ıdac Sur´ıs, Ruoshi Liu, and Carl V ondrick, “Learning the predictability of the future,” in CVPR, 2021, pp. 12607–12617

    2021

  21. [29]

    Hex: Hierarchical emergence exploitation in self-supervised algorithms,

    Kiran Kokilepersaud, Seulgi Kim, Mohit Prab- hushankar, and Ghassan AlRegib, “Hex: Hierarchical emergence exploitation in self-supervised algorithms,” in 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) . IEEE, 2025, pp. 1111– 1121

    2025

  22. [30]

    Gated temporal diffusion for stochastic long-term dense anticipation,

    Olga Zatsarynna, Emad Bahrami, Yazan Abu Farha, Gi- anpiero Francesca, and Juergen Gall, “Gated temporal diffusion for stochastic long-term dense anticipation,” in European Conference on Computer Vision . Springer, 2024, pp. 454–472

    2024

  23. [31]

    Anticipative feature fusion transformer for multi-modal action anticipation,

    Zeyun Zhong, David Schneider, Michael V oit, Rainer Stiefelhagen, and J ¨urgen Beyerer, “Anticipative feature fusion transformer for multi-modal action anticipation,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2023, pp. 6068–6077

    2023

  24. [32]

    Decoupled weight decay regulariza- tion,

    Ilya Loshchilov, “Decoupled weight decay regulariza- tion,” arXiv preprint arXiv:1711.05101, 2017

    2017 arXiv

  25. [33]

    Sgdr: Stochastic gradient descent with warm restarts,

    Ilya Loshchilov and Frank Hutter, “Sgdr: Stochastic gradient descent with warm restarts,” arXiv preprint arXiv:1608.03983, 2016

    2016 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.