pith. sign in

arxiv: 2606.25317 · v1 · pith:DOLL43ORnew · submitted 2026-06-24 · 💻 cs.CV · cs.AI

ESTANet: Efficient Online Error Detection in Procedural Videos via Prediction Inconsistency

Shih-Po Lee , Reza Ghoddoosian , Faizan Siddiqui , Enna Sachdeva , Behzad Dariush This is my paper

Pith reviewed 2026-06-25 21:25 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords error detectionprocedural videosaction detectiononline detectionprediction inconsistencyreal-time inferencelightweight architecture
0
0 comments X

The pith

Errors in procedural videos can be detected online by measuring inconsistencies among predictions from a small set of action detectors.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that action detectors naturally disagree more when a procedure contains errors than when it proceeds correctly. Standard detectors and error-sensitive variants are built to match on normal executions but diverge on mistakes, while detectors that use different lengths of temporal context increase the disagreement further. These mismatches are tallied by majority vote to mark error frames without any extra error-specific labels or heavy model design. The resulting lightweight system runs in real time and reaches state-of-the-art accuracy on three procedural video benchmarks. This approach demonstrates that useful error signals already exist inside ordinary action-prediction behavior.

Core claim

ESTANet detects errors by constructing standard and error-sensitive action detectors that behave similarly on correct executions but respond differently when errors occur, then amplifying inconsistencies with detectors that operate on different temporal contexts, and finally aggregating mismatches through majority voting to flag error frames during online inference.

What carries the argument

ESTANet framework that detects errors by aggregating prediction mismatches between standard, error-sensitive, and temporally varied action detectors via majority voting.

If this is right

  • Online error detection becomes possible with existing action detectors and no specialized supervision.
  • Real-time performance is maintained because the method adds only lightweight comparison and voting steps.
  • The approach applies directly to any procedural video task where action detectors can be run in parallel.
  • Detection accuracy improves when the temporal contexts of the detectors are chosen to differ substantially.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same mismatch principle could be tested on non-procedural anomaly detection tasks such as surveillance or robotics failures.
  • Existing pretrained action models might be reused without retraining simply by varying their input temporal windows.
  • The voting step could be replaced by learned fusion if labeled error data later becomes available.
  • This technique might lower the data requirements for training human-assistance systems that correct user mistakes.

Load-bearing premise

Action detectors will naturally produce sufficiently different predictions on error frames versus correct frames so that majority voting on mismatches can reliably identify errors.

What would settle it

On the EgoPER, Assembly-101-O or EPIC-Tent-O test sets, the standard and error-sensitive detectors produce nearly identical predictions on frames containing known errors, or majority voting fails to mark most of those frames as errors.

Figures

Figures reproduced from arXiv: 2606.25317 by Behzad Dariush, Enna Sachdeva, Faizan Siddiqui, Reza Ghoddoosian, Shih-Po Lee.

Figure 1
Figure 1. Figure 1: The pipeline of our ESTANet. At time t, it produces action predictions y s-st t and y s-lg t by standard detectors and y e-st t and y e-lg t by error-sensitive detectors trained on small and large window sizes (s and l frames). The final error prediction is obtained by majority voting over the four agreement pairs based on action predictions. strategy as in EgoPED for error detection. On the other hand, am… view at source ↗
Figure 2
Figure 2. Figure 2: Pipeline of the TAD module with window size s. The TAD module with window size l uses the same pipeline. 3.2 Standard and Error-Sensitive Action Detector We construct standard action detectors to produce stable action predictions and the sensitive ones to predict inconsistent actions where their inconsistencies capture especially execution errors. We begin with the forward pass of standard action detectors… view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of temporally-varying attribute. The bottom region demonstrates an example where errors occur (missing Action [A1] and doing Error [E]). Remark 2. The design of l enables the action detector depending on learned long-range dependency, specifically, the context spanning complete preceding and partial ongoing actions, when predicting the current action. The strategy ensures our action detectors … view at source ↗
Figure 4
Figure 4. Figure 4: Histograms of frame-wise predicted actions y s-lg (red) and y e-lg (blue) for correct actions (left) and their corresponding execution errors (right) in tea (a) and quesadilla (b) from EgoPER. The x-axis denotes action categories, and the y-axis represents the percentage of frames predicted as each action. while maintaining competitive C-F1 scores (69.5% and 48.0%) on Assembly￾101-O and EPIC-Tent-O, respec… view at source ↗
Figure 5
Figure 5. Figure 5: Each row in a sub-figure, from top to bottom shows frame-wise ground-truth action classes y¯, y s-lg and y e-lg on quesadilla (a) and tea (b) of EgoPER. Each color represents an action class. detection. ESTANet achieves real-time processing (24.4 FPS) in the combined setting, and outperforms PREGO and MistSense, which attain 1.33 FPS with a LLM and 2.5 FPS with a heavier feature extractor and LLM, respecti… view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative visualization of online error detection on EgoPER. Each row from top to bottom shows specific erroneous frames, GT error detection, and error detection predicted by PREGO, DTGL, and ESTANet. 𝑦 𝑠−𝑠𝑡 𝑦 𝑠−𝑙𝑔 Missing step: Scoop Nutella GT Missing step: Measure water Missing step: Add honey Place tortilla on cutting board Sprinkle cinnamon Measure oats Put tea bag into trash Stir using spoon [PITH… view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative visualization of frame-wise predictions when procedural errors (missing steps) occur on EgoPER. numbers of preceding actions, and weaken error sensitivity. In comparison, θ = 50 provides a balanced context of preceding actions, yielding the best performance of 21.6% F1@50 on EgoPER, 59.6% Avg-F1 on Assembly-101-O, and 70.2% Avg-F1 on EPIC-Tent-O. Qualitative Analysis. We visualize the online er… view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative visualization of faliure cases on EgoPER. 5 Limitations In this section, we analyze the failure cases of ESTANet. First, when the dis￾crepancy between the correct action and the error is subtle (leftmost region in [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
read the original abstract

An efficient and accurate system for detecting errors in procedural tasks is crucial for supporting human needs in daily life, as it can provide instant notifications and guide people to correct mistakes. In this work, we study real-time online error detection in procedural videos from a simple but overlooked perspective: the prediction behavior of action detectors themselves. Instead of designing complex architectures or specialized supervision, we observe that action detectors naturally exhibit different prediction characteristics depending on their sensitivity to input dynamics and temporal context. We therefore propose ESTANet (Error-Sensitive and Temporally-vArying Network), a lightweight framework that detects errors by exploiting inconsistencies among action predictions produced by a small set of action detectors. We construct standard and error-sensitive action detectors that behave similarly on correct executions but respond differently when errors occur. Meanwhile, detectors operating with different temporal contexts further amplify prediction inconsistencies when the procedure deviates from the intended sequence. During inference, we detect errors by aggregating mismatches between standard and error-sensitive predictions through majority voting to flag frames that contain errors. Extensive experiments on EgoPER, Assembly-101-O, and EPIC-Tent-O demonstrate that ESTANet achieves state-of-the-art performance in online error detection while maintaining real-time efficiency with a lightweight architecture. Our results highlight that leveraging the intrinsic properties of action detectors can yield a powerful and practical solution for online error detection without increasing architectural design complexity.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes ESTANet, a lightweight framework for real-time online error detection in procedural videos. It constructs standard and error-sensitive action detectors (with differing temporal contexts) that are claimed to behave similarly on correct executions but produce inconsistent predictions on errors; errors are then flagged via majority-vote aggregation of prediction mismatches. The method is presented as requiring no specialized supervision or complex architectures, and extensive experiments are said to demonstrate state-of-the-art performance on EgoPER, Assembly-101-O, and EPIC-Tent-O while maintaining real-time efficiency.

Significance. If the central claim holds—that reliable error detection emerges from intrinsic prediction inconsistencies among unmodified or lightly varied action detectors without any error-specific tuning or labels—the result would be significant for practical deployment in assistive systems, as it avoids the cost of error-labeled data and heavy models. The approach also offers a falsifiable test of whether detector sensitivity differences can be leveraged in an unsupervised manner for procedural tasks.

major comments (2)
  1. [Abstract] Abstract: the claim of 'no specialized supervision' and that detectors 'naturally exhibit different prediction characteristics' is load-bearing for the entire contribution, yet the construction of the error-sensitive variant is not shown to be free of indirect error-data influence (e.g., hyperparameter selection or architecture choice validated on error-containing sequences). If any such validation occurred, the inconsistency signal becomes supervised and the majority-voting step no longer demonstrates unsupervised leverage.
  2. [Abstract] Abstract: SOTA performance is asserted on three datasets, but no quantitative results, error bars, ablation tables, or baseline comparisons are provided, preventing verification that reported gains are robust rather than post-hoc or dataset-specific.
minor comments (1)
  1. [Abstract] The abstract states 'extensive experiments demonstrate SOTA' without any metrics; the full manuscript should include at least one results table with numbers, standard deviations, and runtime measurements to support the efficiency claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. Below we address each major comment point by point with honest responses based on the work presented.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'no specialized supervision' and that detectors 'naturally exhibit different prediction characteristics' is load-bearing for the entire contribution, yet the construction of the error-sensitive variant is not shown to be free of indirect error-data influence (e.g., hyperparameter selection or architecture choice validated on error-containing sequences). If any such validation occurred, the inconsistency signal becomes supervised and the majority-voting step no longer demonstrates unsupervised leverage.

    Authors: The error-sensitive detectors are obtained by applying standard action-recognition backbones with altered temporal context lengths (shorter or longer receptive fields) relative to the standard detectors. These length choices follow well-established principles in the action-recognition literature regarding sensitivity to motion dynamics and are fixed prior to any exposure to the target datasets; no hyperparameter search, architecture selection, or validation step was performed on sequences that contain errors. All training uses only the standard action labels provided by the datasets, with no error annotations or error-specific signals involved at any stage. Consequently, the observed prediction inconsistencies on erroneous frames emerge from the intrinsic differences in temporal sensitivity rather than from any form of error-data supervision. revision: no

  2. Referee: [Abstract] Abstract: SOTA performance is asserted on three datasets, but no quantitative results, error bars, ablation tables, or baseline comparisons are provided, preventing verification that reported gains are robust rather than post-hoc or dataset-specific.

    Authors: The abstract serves as a high-level overview; all quantitative evidence—including per-dataset accuracies, comparisons against published baselines, ablation studies on the number and configuration of detectors, and statistical significance indicators—is reported in full in Section 4, accompanied by Tables 1–4 and Figures 3–6. We agree that embedding one or two headline numbers (e.g., “+4.2 % mAP on EgoPER”) would improve immediate readability of the abstract and will incorporate such figures in the revised version. revision: partial

Circularity Check

0 steps flagged

No significant circularity; empirical aggregation of detector outputs with no derivation chain reducing to fitted inputs

full rationale

The paper describes an empirical framework that constructs standard and error-sensitive detectors (via differing sensitivity to dynamics and temporal context) and aggregates prediction mismatches via majority voting. No equations, parameter fits, or self-citation chains are presented that reduce the claimed error-detection performance to the inputs by construction. The method is self-contained as a direct application of observed detector behaviors on external benchmarks (EgoPER, Assembly-101-O, EPIC-Tent-O), with no load-bearing self-definitional steps or uniqueness theorems imported from prior author work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, no explicit free parameters, axioms, or invented entities are identifiable; the approach relies on the unstated assumption that standard action detectors already encode the necessary sensitivity differences.

pith-pipeline@v0.9.1-grok · 5792 in / 1032 out tokens · 21642 ms · 2026-06-25T21:25:57.097888+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references

  1. [1]

    Princeton University Press, Princeton, NJ (2008)

    Absil, P.A., Mahony, R., Sepulchre, R.: Optimization Algorithms on Matrix Man- ifolds. Princeton University Press, Princeton, NJ (2008)

    2008

  2. [2]

    In: IEEE International Conference on Computer Vision (2023)

    An, J., Kang, H., Han, S.H., Yang, M.H., Kim, S.J.: Miniroad: Minimal rnn frame- work for online action detection. In: IEEE International Conference on Computer Vision (2023)

    2023

  3. [3]

    Neural Information Process- ing Systems (2023)

    Ashutosh, K., Ramakrishnan, S.K., Afouras, T., Grauman, K.: Video-mined task graphs for keystep recognition in instructional videos. Neural Information Process- ing Systems (2023)

    2023

  4. [4]

    Arxiv (2018)

    Battaglia, P.W., Hamrick, J.B., Bapst, V., Sanchez-Gonzalez, A., Zambaldi, V., Malinowski, M., Tacchetti, A., Raposo, D., Santoro, A., Faulkner, R., Gulcehre, C., Song, F., Ballard, A., Gilmer, J., Dahl, G., Vaswani, A., Allen, K., Nash, C., Langston, V., Dyer, C., Heess, N., Wierstra, D., Kohli, P., Botvinick, M., Vinyals, O., Li, Y., Pascanu, R.: Relati...

    2018

  5. [5]

    Bertasius, G., Wang, H., Torresani, L.: Is space-time attention all you need for video understanding? In: Proceedings of the International Conference on Machine Learning (ICML) (July 2021)

    2021

  6. [6]

    arXiv: 2307.16453 (2023)

    Ding, G., Sener, F., Ma, S., Yao, A.: Every mistake counts in assembly. arXiv: 2307.16453 (2023)

    arXiv 2023

  7. [7]

    In: NeurIPS (2021)

    Dvornik, N., Hadji, I., Derpanis, K.G., Garg, A., Jepson, A.D.: Drop-dtw: Aligning common signal between sequences while dropping outliers. In: NeurIPS (2021)

    2021

  8. [8]

    IEEE Conference on Computer Vision and Pattern Recognition (2023)

    Dvornik, N., Hadji, I., Zhang, R., Derpanis, K., Garg, A., Wildes, R., Jepson, A.: Stepformer: Self-supervised step discovery and localization in instructional videos. IEEE Conference on Computer Vision and Pattern Recognition (2023)

    2023

  9. [9]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Eun, H., Moon, J., Park, J., Jung, C., Kim, C.: Learning to discriminate informa- tion for online action detection. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 809–818 (2020)

    2020

  10. [10]

    In: Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition

    Farha, Y.A., Gall, J.: Ms-tcn: Multi-stage temporal convolutional network for ac- tion segmentation. In: Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition. pp. 3575–3584 (2019)

    2019

  11. [11]

    IEEE Conference on Computer Vision and Pattern Recognition (2024)

    Flaborea, A., Melendugno, G., Pliniq, L., Scofanoq, L., Matteisq, E., Furnari, A., Farinella, G., Galasso, F.: Prego: online mistake detection in procedural egocentric videos. IEEE Conference on Computer Vision and Pattern Recognition (2024)

    2024

  12. [12]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Ghoddoosian, R., Dwivedi, I., Agarwal, N., Dariush, B.: Weakly-supervised action segmentation and unseen error detection in anomalous instructional videos. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 10128–10138 (2023)

    2023

  13. [13]

    Lee et al

    Grauman, K., Westbury, A., Byrne, E., Chavis, Z., Furnari, A., Girdhar, R., Ham- burger, J., Jiang, H., Liu, M., Liu, X., Martin, M., Nagarajan, T., Radosavovic, I., Ramakrishnan, S.K., Ryan, F., Sharma, J., Wray, M., Xu, M., Xu, E.Z., Zhao, C., Bansal, S., Batra, D., Cartillier, V., Crane, S., Do, T., Doulaty, M., Era- palli, A., Feichtenhofer, C., Frago...

    2022

  14. [14]

    In: European Conference on Computer Vision (2022)

    Guo, H., Ren, Z., Wu, Y., Hua, G., Ji, Q.: Uncertainty-based spatial-temporal attention for online action detection. In: European Conference on Computer Vision (2022)

    2022

  15. [15]

    In: IEEE Conference on Computer Vision and Pattern Recognition (2025)

    Huang, W.J., Li, Y.M., Xia, Z.W., Tang, Y.M., Lin, K.Y., Hu, J.F., Zheng, W.S.: Modeling multiple normal action representations for error detection in procedural tasks. In: IEEE Conference on Computer Vision and Pattern Recognition (2025)

    2025

  16. [16]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Huang, Y., Chen, G., Xu, J., Zhang, M., Yang, L., Pei, B., Zhang, H., Dong, L., Wang, Y., Wang, L., Qiao, Y.: Egoexolearn: A dataset for bridging asynchronous ego- and exo-centric view of procedural activities in real world. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 22072–22086 (June 2024)

    2024

  17. [17]

    International Conference on Computer Vision Workshop (2019)

    Jang, Y., Sullivan, B., Ludwig, C., Gilchrist, I., Damen, D., Mayol-Cuevas, W.: Epictent: An egocentric video dataset for camping tent assembly. International Conference on Computer Vision Workshop (2019)

    2019

  18. [18]

    ICLR 2023 Workshop on Multimodal Representation Learning: Perks and Pitfalls (2023)

    Jang, Y., Sohn, S., Logeswaran, L., Luo, T., Lee, M., Lee, H.: Multimodal subtask graph generation from instructional videos. ICLR 2023 Workshop on Multimodal Representation Learning: Perks and Pitfalls (2023)

    2023

  19. [19]

    IEEE Conference on Computer Vision and Pattern Recognition (2024)

    Lee, S., Lu, Z., Zhang, Z., Hoai, M., Elhamifar, E.: Error detection in egocen- tric procedural task videos. IEEE Conference on Computer Vision and Pattern Recognition (2024)

    2024

  20. [20]

    In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (jun 2022)

    Li, M., Chen, L., Duarr, Y., Hu, Z., Feng, J., Zhou, J., Lu, J.: Bridge-prompt: Towards ordinal action understanding in instructional videos. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (jun 2022)

    2022

  21. [21]

    IEEE Transactions on Pattern Analysis and Machine Intelligence pp

    Li, S.J., AbuFarha, Y., Liu, Y., Cheng, M.M., Gall, J.: Ms-tcn++: Multi-stage temporal convolutional network for action segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence pp. 1–1 (2020).https://doi.org/10. 1109/TPAMI.2020.3021756

    arXiv 2020

  22. [22]

    European Conference on Computer Vision (2024)

    Li, Z., Chen, Q., Han, T., Zhan, Y., Wang, Y., Xie, W.: Multi-sentence grounding for long-term instructional video. European Conference on Computer Vision (2024)

    2024

  23. [23]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Liu, Y., Huo, J., Peng, J., Sparks, R., Dasgupta, P., Granados, A., Ourselin, S.: Skit: a fast key information video transformer for online surgical phase recognition. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 21074–21084 (2023)

    2023

  24. [24]

    IEEE Conference on Computer Vision and Pattern Recognition (2024)

    Lu, Z., Elhamifar, E.: Fact: Frame-action cross-attention temporal modeling for efficient action segmentation. IEEE Conference on Computer Vision and Pattern Recognition (2024)

    2024

  25. [25]

    Inter- national Conference on Computer Vision (2025)

    Lu, Z., Elhamifar, E.: Multi-modal few-shot temporal action segmentation. Inter- national Conference on Computer Vision (2025)

    2025

  26. [26]

    IEEE Conference on Computer Vision and Pattern Recognition (2025)

    Lu,Z.,Iftekhar,A.,Mittal,G.,Meng,T.,Wang,X.,Zhao,C.,Kukkala,R.,Elhami- far, E., Chen, M.: Decafnet: Delegate and conquer for efficient temporal grounding in long videos. IEEE Conference on Computer Vision and Pattern Recognition (2025)

    2025

  27. [27]

    Neural Information Processing Systems (2024) ESTANet 17

    Luigi Seminara, Giovanni Maria Farinella, A.F.: Differentiable task graph learning: Procedural activity representation and online mistake detection from egocentric videos. Neural Information Processing Systems (2024) ESTANet 17

    2024

  28. [28]

    IEEE Con- ference on Computer Vision and Pattern Recognition (2024)

    Mu, F., Mo, S., Li, Y.: Snag: Scalable and accurate video grounding. IEEE Con- ference on Computer Vision and Pattern Recognition (2024)

    2024

  29. [29]

    In: IEEE Conference on Computer Vision and Pattern Recognition (2025)

    Pang, Z., Sener, F., Yao, A.: Context-enhanced memory-refined transformer for online action detection. In: IEEE Conference on Computer Vision and Pattern Recognition (2025)

    2025

  30. [30]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    Patsch, C., Wu, Y., Zakour, M., Salihu, D., Steinbach, E.: Mistsense: Versatile online detection of procedural and execution mistakes. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2025)

    2025

  31. [31]

    Peddi, R., Arya, S., Challa, B., Pallapothula, L., Vyas, A., Gouripeddi, B., Wang, J., Zhang, Q., Komaragiri, V., Ragan, E., Ruozzi, N., Xiang, Y., Gogate, V.: Cap- tainCook4D: A Dataset for Understanding Errors in Procedural Activities (2024), https://arxiv.org/abs/2312.14556

    arXiv 2024

  32. [32]

    In: Neural Information Processing Systems (2024)

    Seminara, L., Farinella, G.M., Furnari, A.: Differentiable task graph learning: Pro- ceduralactivityrepresentationandonlinemistakedetectionfromegocentricvideos. In: Neural Information Processing Systems (2024)

    2024

  33. [33]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Sener, F., Chatterjee, D., Shelepov, D., He, K., Singhania, D., Wang, R., Yao, A.: Assembly101: A large-scale multi-view video dataset for understanding procedural activities. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 21096–21106 (2022)

    2022

  34. [34]

    IEEE Conference on Computer Vision and Pattern Recognition (2024)

    Shen, Y., Elhamifar, E.: Progress-aware online action segmentation for egocen- tric procedural task videos. IEEE Conference on Computer Vision and Pattern Recognition (2024)

    2024

  35. [35]

    International Conference on Learning Represen- tations (2020)

    Sohn, S., Woo, H., Choi, J., Lee, H.: Meta reinforcement learning with autonomous inference of subtask dependencies. International Conference on Learning Represen- tations (2020)

    2020

  36. [36]

    PAMI (2021)

    Souri,Y.,Fayyaz,M.,Minciullo,L.,Francesca,G.,Gall,J.:FastWeaklySupervised Action Segmentation Using Mutual Consistency. PAMI (2021)

    2021

  37. [37]

    In: Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision

    Wang, J., Chen, G., Huang, Y., Wang, L., Lu, T.: Memory-and-anticipation trans- former for online action understanding. In: Proceedings of the IEEE/CVF Inter- national Conference on Computer Vision. pp. 13824–13835 (2023)

    2023

  38. [38]

    Wang, X., Zhang, S., Qing, Z., Shao, Y., Zuo, Z., Gao, C., Sang, N.: Oadtr: Online actiondetectionwithtransformers.In:ProceedingsoftheIEEE/CVFInternational Conference on Computer Vision. pp. 7565–7575 (2021)

    2021

  39. [39]

    IEEE International Conference on Computer Vision (2023)

    Wang, X., Kwon, T., Pan, M.R.B., Chakraborty, I., Andrist, S.: Holoassist: an egocentric human interaction dataset for interactive ai assistants in the real world. IEEE International Conference on Computer Vision (2023)

    2023

  40. [40]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Xu, M., Gao, M., Chen, Y.T., Davis, L.S., Crandall, D.J.: Temporal recurrent net- works for online action detection. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 5532–5541 (2019)

    2019

  41. [41]

    In: Advances in Neural Information Processing Systems (2021)

    Xu, M., Xiong, Y., Chen, H., Li, X., Xia, W., Tu, Z., Soatto, S.: Long short- term transformer for online action detection. In: Advances in Neural Information Processing Systems (2021)

    2021

  42. [42]

    arXiv (2024)

    Yang, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., Li, C., Liu, D., Huang, F., Wei, H., Lin, H., Yang, J., Tu, J., Zhang, J., Yang, J., Yang, J., Zhou, J., Lin, J., Dang, K., Lu, K., Bao, K., Yang, K., Yu, L., Li, M., Xue, M., Zhang, P., Zhu, Q., Men, R., Lin, R., Li, T., Tang, T., Xia, T., Ren, X., Ren, X., Fan, Y., Su, Y., Zhang, Y., Wan, Y., Li...

    2024

  43. [43]

    In: The British Machine Vision Conference (BMVC) (2021) 18 S.-P

    Yi, F., Wen, H., Jiang, T.: Asformer: Transformer for action segmentation. In: The British Machine Vision Conference (BMVC) (2021) 18 S.-P. Lee et al

    2021

  44. [44]

    In: European Conference on Computer Vision

    Zhao, Y., Krähenbühl, P.: Real-time online video detection with temporal smooth- ing transformers. In: European Conference on Computer Vision. pp. 485–502. Springer (2022)

    2022