Pith. sign in

REVIEW 5 major objections 5 minor 51 references

Video Domain Incremental Learning for Human Action Recognition in Home Environments

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

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

Pith's one-line read The paper formalizes Video Domain Incremental Learning for home action recognition and shows that a simple replay baseline—reservoir sampling plus knowledge distillation, no domain labels—outperforms most existing continual learning…

desk verdict Useful VDIL benchmark idea, but the Toyota scene split likely breaks the fixed-class assumption and several internal inconsistencies need fixing before this is a reliable community resource. read the letter →

arxiv 2412.16946 v1 pith:V4BYACPB submitted 2024-12-22 cs.CV

Yuanda Hu , Xing Liu , Meiying Li , Yate Ge , Xiaohua Sun , Weiwei Guo This is my paper · ORCID
classification cs.CV
keywords videodomainincrementallearningactionrecognitioncontinualcatastrophicforgettingreplaybufferreservoirsamplinghomeenvironmentsshift
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

Video Domain Incremental Learning (VDIL) is the problem of adapting an action-recognition model to a sequence of new domains—different users, rooms, or camera views—while the set of action classes stays fixed. This paper argues that home environments produce exactly such shifts, that they cause catastrophic forgetting, and that the problem has been overlooked relative to class-incremental video learning. To test this, the paper builds three benchmarks by splitting NTU RGB+D by user, Toyota Smarthome by scene, and ETRI-Activity3D-LivingLab by both. It then proposes DRIFT, a plain replay baseline that fills a memory buffer with reservoir sampling and trains with classification plus knowledge-distillation losses, requiring no domain labels. The central claim is that this simple strategy outperforms most existing continual learning methods across the three benchmarks, with the largest margin on the scene-based split.

What carries the argument

DRIFT (Domain Replay Incremental Feature Training) is the central mechanism. Its key parts are reservoir sampling, which keeps the memory buffer a fair sample of all domains seen so far without needing to know where one domain ends and another begins, and a dual loss: a classification cross-entropy term over current and buffered samples, plus a knowledge-distillation term that pins the current model's softened outputs to the previous model's outputs on new data. The benchmark machinery consists of domain splits (user, scene, hybrid) over three RGB-D datasets with a fixed class set, evaluated by average accuracy and backward forgetting.

What would settle it

Compute, for each domain task in NTU RGB+D, Toyota Smarthome, and ETRI-Activity3D-LivingLab, the set of action classes actually present. If any task lacks a substantial class, the average-accuracy and backward-forgetting numbers conflate domain shift with class imbalance, and the benchmark does not measure what it claims. Resolving the 31-versus-55 class-count discrepancy for Toyota Smarthome would be a concrete first check.

Watch

Extended reading notes

Core claim

In home action recognition, the dominant source of representation drift is the domain itself, not the action class: the paper's t-SNE visualization of Toyota Smarthome shows samples grouping by camera viewpoint rather than by activity. The paper therefore formalizes VDIL and constructs a benchmark where each task is one domain and every task shares the full class set, with average accuracy and backward forgetting as metrics. The proposed DRIFT baseline—reservoir sampling into a fixed memory plus cross-entropy on current and replayed data and knowledge distillation from the previous model—reaches 78.45% average accuracy on the scene-based split, ahead of all compared methods, and stays competitive on the user and hybrid splits while using no task identifiers. The accompanying ablations show the two losses are complementary and that random sample selection outperforms entropy-based and class-prototype selection for this setting.

Load-bearing premise

The benchmark assumes every task (domain) contains all action classes in sufficient number, so that the accuracy metrics measure pure domain shift; the paper states this assumption but never verifies it on the three datasets, and its own text gives contradictory class counts for Toyota Smarthome (31 activities in Section 3.1 versus 55 classes in Table 1).

Editorial extensions

If this is right

  • A task-agnostic replay method with no domain labels is a viable default baseline for video domain-incremental learning, not a stopgap.
  • Scene-based domain shifts in home environments are harder than user-based shifts, so deployment systems should prioritize viewpoint and room diversity when collecting data.
  • Memory-constrained setups can use reservoir sampling: even a 225-sample buffer beats larger buffers of more complex methods on the scene benchmark.
  • The dual-loss recipe (cross-entropy on replay plus knowledge distillation) transfers directly to other video continual-learning settings with fixed classes.

Reading between the lines

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

  • The result suggests that for many real-world video deployments, sophisticated regularizers or task-boundary-aware buffers may be unnecessary; the bottleneck is storing representative old samples.
  • One could test the same benchmark protocol on egocentric or surveillance video, where domain shifts are driven by wearer or viewpoint rather than home scenes.
  • The benchmark's value depends on verifying that every split indeed contains all classes with enough samples; the paper does not show per-task class distributions, so the reported accuracy and forgetting numbers should be read with that caveat until the protocol is checked.
  • Future VDIL work could combine reservoir replay with prompt-based or parameter-efficient adapters, potentially closing the remaining gap to joint training while keeping the task-agnostic property.
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

5 major / 5 minor

Summary. The paper formalizes Video Domain Incremental Learning (VDIL) for human action recognition in home environments, a continual-learning setting in which the action-class set stays fixed while the data distribution shifts across tasks. It introduces three benchmarks built from public datasets by splitting NTU RGB+D into user domains, Toyota Smarthome into scene domains, and ETRI-Activity3D-LivingLab into hybrid user-scene domains. The authors propose DRIFT, a replay-and-distillation baseline that uses reservoir sampling and a dual classification-plus-KD loss, and compare it against Naive, Joint, LwF, EWC, DER, iCaRL, and BiC on average accuracy (AA) and backward forgetting (BWF). The reported results show that DRIFT outperforms all non-replay baselines and most replay baselines, with particularly large gains on Toyota Smarthome, and the paper includes ablations on loss components, sampling strategies, and memory size.

Significance. If the benchmark protocol is valid, this would be one of the first dedicated video domain-incremental learning benchmarks, addressing a real gap since most video continual-learning work targets class-incremental settings. The three split types are a sensible way to isolate user-driven, scene-driven, and combined domain shift, and the choice of widely used public action-recognition datasets makes the benchmark accessible. DRIFT is a simple and competitive task-agnostic baseline; on Toyota Smarthome it improves substantially over the compared methods, and its buffer-size analysis is a useful starting point for resource-constrained deployment. However, the central claim depends on the assumption that every task contains the full fixed class set, and the paper contains internal inconsistencies that call this assumption into question. The benchmark and the comparative results are potentially valuable, but the protocol must be verified and the reported numbers reconciled before the claims can be accepted.

major comments (5)
  1. [Sec. 3 and Table 1] The problem definition in Sec. 3 states that each task Dk encompasses all classes in C, but the paper never verifies this for any of the three splits, and Table 1 lists Toyota Smarthome as having 55 classes while Sec. 3.1 states the dataset has 31 activities. Because Toyota Smarthome activities are naturally room-specific, scene-domain splits may omit many classes from each task; if so, the AA and BWF metrics measure partial-domain or class-imbalanced learning rather than pure domain shift. Please report per-task class coverage and class-frequency statistics for all three benchmarks, and either restrict the evaluation to classes that appear in every task or explicitly justify why the fixed-class-set protocol is preserved.
  2. [Sec. 4.3 and Table 3] The buffer-size section states that a memory size of 550 achieves 83.17% and 'exceeds' Joint, but Table 3 shows that 83.17 is obtained at memory size 1100, while the 550 row reports 78.45. The same paragraph says 550 corresponds to 15.6% of the training set, which does not match 550/8,848 ≈ 6.2%. This inconsistency directly affects the paper's claim about matching or exceeding the non-incremental upper bound; please correct the table or the text and recompute the stated percentage.
  3. [Sec. 4.1 and Table 2] The main comparison in Table 2 does not specify the memory buffer size used for each method, and Sec. 4.1 only mentions method-specific hyperparameters. Without knowing whether all replay methods use the same buffer size and the same buffer-management rules, the headline claim that DRIFT outperforms most existing continual learning methods cannot be assessed fairly. Please state the buffer size used for DRIFT, iCaRL, DER, and BiC in the main experiments, and clarify how each method's buffer is populated.
  4. [Sec. 3.2, Eq. (5)] The knowledge-distillation loss in Eq. (5) is written as a cross-entropy between f_theta_{t-1}(V)/T and f_theta_t(V)/T, but dividing raw network outputs by T does not produce a valid probability distribution unless a softmax is applied to the logits before the temperature scaling. As written, the loss is not a standard distillation objective and its behavior is undefined for unbounded logits. Please clarify whether f denotes logits or probabilities, use softmax(z/T) consistently, and also explain why distillation is applied only to current-task samples and not to memory-buffer samples.
  5. [Sec. 3.2 and Sec. 4.2] The paper emphasizes that DRIFT is task-agnostic and 'does not require knowledge of task identifiers', but the evaluation protocol is not fully specified: the metrics ak,T in Eqs. (2)-(3) require evaluating each task separately, and it is unclear whether task identity is available at test time for any method. Please state explicitly whether test samples are evaluated with or without task IDs, and whether the iCaRL and BiC baselines are given task boundaries during training. This is essential for judging the claimed advantage of DRIFT in task-agnostic scenarios.
minor comments (5)
  1. [Table 1] The word 'Hybird' in the last row of Table 1 should be 'Hybrid'.
  2. [Table 2] The table header contains the typo 'ackward Transfer' instead of 'Backward Transfer', and the text in Sec. 4.3 uses both BWF and BWT for the same metric; please standardize the terminology.
  3. [Fig. 1 caption] The caption says the bar plots show catastrophic forgetting, but the figure appears to contain line plots or charts rather than bar plots; please align the caption with the actual figure.
  4. [Sec. 4.3] The sentence describing the 550-memory result should be rewritten to match Table 3: the 83.17% value corresponds to memory size 1100, and the 550 value is 78.45%.
  5. [Sec. 4.1] Please state whether the TSN backbone is initialized with ImageNet-pretrained weights for all compared methods, and whether any method uses additional data augmentation besides the DER horizontal flips mentioned in the text, so that the comparison is fully reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's claims are empirical benchmark results with external baselines; the benchmark protocol inconsistency is a validity risk, not a circular-reasoning defect.

full rationale

I walked the derivation chain. The paper's central claim is that DRIFT, a reservoir-sampling-plus-distillation baseline, outperforms most existing continual learning methods on three proposed VDIL benchmarks. This is an empirical claim evaluated against named external methods (LwF, EWC, iCaRL, BiC, DER) and against Joint/Naive bounds; none of these comparisons reduces to a fitted parameter or to the method's own definition. The problem formulation (Sec. 3) defines AA and BWF from accuracies a_{k,T}, and the proposed loss (Eq. 6) combines cross-entropy with KD; no parameter is fit to the benchmark and then reported as a prediction. There are no self-citations by the authors that carry a load-bearing argument, no imported uniqueness theorem, and no ansatz smuggled in via citation. The one substantive concern is that the benchmark may not satisfy the stated protocol that every task contains all classes C: Sec. 3.1 says Toyota Smarthome has 31 activities while Table 1 lists 55 classes, and scene-split tasks may have incomplete class coverage. That is a correctness/validity threat to the benchmark claim, but it is not circular reasoning, because the method's comparison does not define its success metric in terms of the benchmark construction. Score 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The benchmark and baseline rest on standard deep learning assumptions plus an unverified protocol assumption that every domain task contains all classes with adequate samples. The paper does not report the KD loss weight, so the main result is not fully specified. No invented physical or conceptual entities are introduced.

free parameters (3)
  • KD loss weight lambda = not reported
    Appears in Eq. 6 as the balance between classification and distillation losses. No value is given anywhere in the experiments, and the ablation in Table 4 shows the combined loss drives the 78.45% result.
  • Memory buffer size = 550 (also 1100 and 225)
    The replay buffer size is central to the proposed task-agnostic strategy and the buffer analysis in Sec. 4.3. Results change substantially with buffer size, from 83.17 at 1100 to 74.71 at 225.
  • Distillation temperature T = 2
    The knowledge distillation loss in Eq. 5 uses a fixed softmax temperature of 2, chosen by hand without a sweep or sensitivity analysis.
assumptions (4)
  • domain assumption Each task Dk contains all action classes C with sufficient samples.
    Stated in Sec. 3 and Eq. 1 so that AA and BWF measure fixed-class domain shift. Never verified for NTU (40 users), Toyota (7 scenes), or ETRI (40 hybrid domains), and the Toyota class-count inconsistency makes it questionable.
  • domain assumption Pre-trained ImageNet features transfer to home environment RGB video.
    All experiments use a TSN model with ImageNet-pretrained ResNet-50. No from-scratch training or alternative pretraining is compared, so all reported numbers depend on this transfer.
  • domain assumption Reservoir sampling without task boundaries yields an unbiased representation of past domains.
    The task-agnostic claim of DRIFT relies on this property, but no statistical analysis or comparison of buffer composition is provided.
  • domain assumption Domain splits by user, scene, and hybrid are correctly assigned and mutually exclusive.
    If users or camera views overlap between train and test domains, the reported forgetting numbers would not measure true domain incremental learning.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Video Domain Incremental Learning for Human Action Recognition in Home Environments." pith.science (2026). https://pith.science/paper/V4BYACPB

@misc{pith2026241216946,
  author       = {Pith},
  title        = {Pith review of: Video Domain Incremental Learning for Human Action Recognition in Home Environments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V4BYACPB}},
  note         = {Machine review of arXiv:2412.16946}
}
read the original abstract

It is significantly challenging to recognize daily human actions in homes due to the diversity and dynamic changes in unconstrained home environments. It spurs the need to continually adapt to various users and scenes. Fine-tuning current video understanding models on newly encountered domains often leads to catastrophic forgetting, where the models lose their ability to perform well on previously learned scenarios. To address this issue, we formalize the problem of Video Domain Incremental Learning (VDIL), which enables models to learn continually from different domains while maintaining a fixed set of action classes. Existing continual learning research primarily focuses on class-incremental learning, while the domain incremental learning has been largely overlooked in video understanding. In this work, we introduce a novel benchmark of domain incremental human action recognition for unconstrained home environments. We design three domain split types (user, scene, hybrid) to systematically assess the challenges posed by domain shifts in real-world home settings. Furthermore, we propose a baseline learning strategy based on replay and reservoir sampling techniques without domain labels to handle scenarios with limited memory and task agnosticism. Extensive experimental results demonstrate that our simple sampling and replay strategy outperforms most existing continual learning methods across the three proposed benchmarks.

Figures

Figures reproduced from arXiv: 2412.16946 by the authors.

Figure 1
Figure 1. (a) Comparison of Video Class Incremental Learning (VCIL) and Video Domain Incremental Learning (VDIL). VCIL learns new classes incrementally with a fixed data distribution, while VDIL learns from evolving data distributions within a constant class set. (b) The bar plots show the presence of catastrophic forgetting in VDIL paradigm, indicated by the performance degradation on previous tasks when learning unseen doma… view at source ↗
Figure 2
Figure 2. Illustration of the Video Domain Incremental Learning (VDIL) benchmarks, where models learn from evolving data distributions while the action classes re￾main fixed. We split three datasets (NTU RGB+D, Toyota Smarthome, and ETRI￾Activity3D-LivingLab) into different domain types: user, scene, and hybrid (combining user and scene), to create benchmarks that evaluate models’ ability to incrementally adapt to domain shif… view at source ↗
Figure 3
Figure 3. T-SNE visualization of feature distributions in the Toyota Smarthome dataset [20]. Colors represent distinct Scene Domains, each corresponding to different camera viewpoints. Contrary to expectations, samples cluster by domain rather than action class [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: DRIFT leverages reservoir sampling to address the task-agnostic challenge, en￾suring that each domain is equally likely to be represented in the memory. Furthermore, DRIFT incorporates a dual-loss strategy to maintain robust adaptability and memory retention across var…
Figure 5
Figure 5. Figure 5: Performance comparison of continual learning methods across three proposed benchmarks. Panels (a), (b), and (c) illustrate the performance dynamics and final average accuracies of various algorithms on the NTU RGB+D, Toyota Smarthome, and ETRI-Activity3D-LivingLab data…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 37 canonical work pages

  1. [1]

    Just a Glimpse: Rethinking Temporal Information for Video Continual Learning,

    L. Alssum, J. León Alcázar, M. Ramazanova, C. Zhao, and B. Ghanem, “Just a Glimpse: Rethinking Temporal Information for Video Continual Learning,” in2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Vancouver, BC, Canada: IEEE, 2023, pp. 2474–2483. 2, 4

    work page 2023

  2. [2]

    A continual learning survey: Defying forgetting in classification tasks,

    M. Delange, R. Aljundi, M. Masana, S. Parisot, X. Jia, A. Leonardis, G. Slabaugh, and T. Tuytelaars, “A continual learning survey: Defying forgetting in classification tasks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1–1,

    work page

  3. [3]

    Multi-Domain Incremental Learning for Semantic Segmentation,

    P. Garg, R. Saluja, V. N. Balasubramanian, C. Arora, A. Subramanian, and C. Jawahar, “Multi-Domain Incremental Learning for Semantic Segmentation,” in 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). Waikoloa, HI, USA: IEEE, 2022, pp. 2080–2090. 4

    work page 2022

  4. [4]

    An Efficient Domain-Incremental Learning Approach to Drive in All Weather Conditions,

    M. Jehanzeb Mirza, M. Masana, H. Possegger, and H. Bischof, “An Efficient Domain-Incremental Learning Approach to Drive in All Weather Conditions,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Work- shops (CVPRW). New Orleans, LA, USA: IEEE, 2022, pp. 3000–3010. 4

    work page 2022

  5. [5]

    Clustering-based Domain-Incremental Learning,

    C. Lamers, R. Vidal, N. Belbachir, N. Van Stein, T. Bäck, and P. Giampouras, “Clustering-based Domain-Incremental Learning,” in 2023 IEEE/CVF Interna- tional Conference on Computer Vision Workshops (ICCVW). Paris, France: IEEE, 2023, pp. 3376–3384. 4

    work page 2023

  6. [6]

    Class-Incremental Learning: Survey and Performance Evaluation on Image Classification,

    M.Masana,X.Liu,B.Twardowski,M.Menta,A.D.Bagdanov,andJ.VanDeWei- jer, “Class-Incremental Learning: Survey and Performance Evaluation on Image Classification,” IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1–20, 2022. 3, 4

    work page 2022

  7. [7]

    Adversarial Cross-Domain Action Recognition with Co-Attention,

    B. Pan, Z. Cao, E. Adeli, and J. C. Niebles, “Adversarial Cross-Domain Action Recognition with Co-Attention,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 07, pp. 11815–11822, 2020. 5

    work page 2020

  8. [8]

    Class-Incremental Learning for Action Recogni- tion in Videos,

    J. Park, M. Kang, and B. Han, “Class-Incremental Learning for Action Recogni- tion in Videos,” in2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, QC, Canada: IEEE, 2021, pp. 13678–13687. 4 16 Yuanda Hu et al

    work page 2021

Show all 51 references
  1. [9]

    Learning a Condensed Frame for Memory-Efficient Video Class- Incremental Learning,

    Y. Pei, Z. Qing, J. Cen, X. Wang, S. Zhang, Y. Wang, M. Tang, N. Sang, and X. Qian, “Learning a Condensed Frame for Memory-Efficient Video Class- Incremental Learning,” Advances in Neural Information Processing Systems , vol. 35, pp. 31002–31016, 2022. 4

    2022

  2. [10]

    Space-time Prompting for Video Class-incremental Learning,

    Y. Pei, Z. Qing, S. Zhang, X. Wang, Y. Zhang, D. Zhao, and X. Qian, “Space-time Prompting for Video Class-incremental Learning,” in2023 IEEE/CVF Interna- tional Conference on Computer Vision (ICCV). Paris, France: IEEE, 2023, pp. 11898–11908. 4

    2023

  3. [11]

    A Unified Approach to Domain Incremental Learning with Memory: Theory and Algorithm,

    H. Shi and H. Wang, “A Unified Approach to Domain Incremental Learning with Memory: Theory and Algorithm,”Advances in Neural Information Processing Sys- tems, vol. 36, pp. 15027–15059, 2023. 4

    2023

  4. [12]

    Spatio-temporal Contrastive Domain Adaptation for Action Recognition,

    X. Song, S. Zhao, J. Yang, H. Yue, P. Xu, R. Hu, and H. Chai, “Spatio-temporal Contrastive Domain Adaptation for Action Recognition,” in2021 IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR). Nashville, TN, USA: IEEE, 2021, pp. 9782–9790. 5

    2021

  5. [13]

    Dual-Head Contrastive Domain Adaptation for Video Action Recog- nition,

    V. G. Turrisi Da Costa, G. Zara, P. Rota, T. Oliveira-Santos, N. Sebe, V. Murino, and E. Ricci, “Dual-Head Contrastive Domain Adaptation for Video Action Recog- nition,” in 2022 IEEE/CVF Winter Conference on Applications of Computer Vi- sion (WACV). Waikoloa, HI, USA: IEEE, 20...

    2022

  6. [14]

    PIVOT: Prompting for Video Continual Learning,

    A. Villa, J. L. Alcázar, M. Alfarra, K. Alhamoud, J. Hurtado, F. C. Heilbron, A. Soto, and B. Ghanem, “PIVOT: Prompting for Video Continual Learning,” in 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Vancouver, BC, Canada: IEEE, 2023, pp. 24214–24223. 4

    2023

  7. [15]

    vCLIMB: A Novel Video Class Incremental Learning Benchmark,

    A. Villa, K. Alhamoud, V. Escorcia, F. C. Heilbron, J. L. Alcazar, and B. Ghanem, “vCLIMB: A Novel Video Class Incremental Learning Benchmark,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA: IEEE, 2022, pp. 19013–19022. 4, 10

    2022

  8. [16]

    Confidence Attention and Generalization Enhanced Distillation for Continuous Video Domain Adaptation,

    X. Wang, Y. Xu, J. Yang, B. Wen, and A. C. Kot, “Confidence Attention and Generalization Enhanced Distillation for Continuous Video Domain Adaptation,”

  9. [17]

    S-Prompts Learning with Pre-trained Trans- formers: An Occam’s Razor for Domain Incremental Learning,

    Y. Wang, Z. Huang, and X. Hong, “S-Prompts Learning with Pre-trained Trans- formers: An Occam’s Razor for Domain Incremental Learning,”Advances in Neural Information Processing Systems, vol. 35, pp. 5682–5695, 2022. 4

    2022

  10. [18]

    Interact before Align: Leverag- ing Cross-Modal Knowledge for Domain Adaptive Action Recognition,

    L. Yang, Y. Huang, Y. Sugano, and Y. Sato, “Interact before Align: Leverag- ing Cross-Modal Knowledge for Domain Adaptive Action Recognition,” in2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA: IEEE, 2022, pp. 14702–14712. 5

    2022

  11. [19]

    When Video Classification Meets Incremental Classes,

    H. Zhao, X. Qin, S. Su, Y. Fu, Z. Lin, and X. Li, “When Video Classification Meets Incremental Classes,” inProceedings of the 29th ACM International Conference on Multimedia. Virtual Event China: ACM, 2021, pp. 880–889. 4

    2021

  12. [20]

    Toyota smarthome: Real-world activities of daily living,

    S. Das, R. Dai, M. Koperski, L. Minciullo, L. Garattoni, F. Brémond, and G. Francesca, “Toyota smarthome: Real-world activities of daily living,” in2019 IEEE/CVF International Conference on Computer Vision, ICCV 2019, Seoul, Ko- rea (South), October 27 - November 2, 2019. IEEE...

    2019

  13. [21]

    Three types of incremental learning,

    G. M. Van De Ven, T. Tuytelaars, and A. S. Tolias, “Three types of incremental learning,” Nature Machine Intelligence, vol. 4, no. 12, pp. 1185–1197, 2022. 3, 4

    2022

  14. [22]

    iCaRL: Incremental Classifier and Representation Learning,

    S.-A. Rebuffi, A. Kolesnikov, G. Sperl, and C. H. Lampert, “iCaRL: Incremental Classifier and Representation Learning,” in2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI: IEEE, 2017, pp. 5533–

    2017

  15. [23]

    Dark Experience for General Continual Learning: A Strong, Simple Baseline,

    P. Buzzega, M. Boschini, A. Porrello, D. Abati, and SIMONE. CALDERARA, “Dark Experience for General Continual Learning: A Strong, Simple Baseline,” in Advances in Neural Information Processing Systems, vol. 33. Curran Associates, Inc., 2020, pp. 15920–15930. 3, 4, 10, 12

    2020

  16. [24]

    On Tiny Episodic Memories in Continual Learning,

    A. Chaudhry, M. Rohrbach, M. Elhoseiny, T. Ajanthan, P. K. Dokania, P. H. S. Torr, and M. Ranzato, “On Tiny Episodic Memories in Continual Learning,” 2019. 4

    2019

  17. [25]

    Overcoming catastrophic forgetting in neural net- works,

    J. Kirkpatrick, R. Pascanu, N. Rabinowitz, J. Veness, G. Desjardins, A. A. Rusu, K. Milan, J. Quan, T. Ramalho, A. Grabska-Barwinska, D. Hassabis, C. Clopath, D. Kumaran, and R. Hadsell, “Overcoming catastrophic forgetting in neural net- works,” Proceedings of the National Aca...

    2017

  18. [26]

    Continual Learning with Deep Generative Replay,

    H. Shin, J. K. Lee, J. Kim, and J. Kim, “Continual Learning with Deep Generative Replay,” inAdvances in Neural Information Processing Systems, vol. 30. Curran Associates, Inc., 2017. 4

    2017

  19. [27]

    Learning without Forgetting,

    Z. Li and D. Hoiem, “Learning without Forgetting,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 40, no. 12, pp. 2935–2947, 2018. 5, 8, 10, 12

    2018

  20. [28]

    Large Scale Incremen- tal Learning,

    Y. Wu, Y. Chen, L. Wang, Y. Ye, Z. Liu, Y. Guo, and Y. Fu, “Large Scale Incremen- tal Learning,” in 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, CA, USA: IEEE, 2019, pp. 374–382. 5, 10, 12, 13

    2019

  21. [29]

    PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning,

    A. Mallya and S. Lazebnik, “PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 7765–7773. 5

    2018

  22. [30]

    Piggyback: Adapting a Single Network to Multiple Tasks by Learning to Mask Weights,

    A. Mallya, D. Davis, and S. Lazebnik, “Piggyback: Adapting a Single Network to Multiple Tasks by Learning to Mask Weights,” inProceedings of the European Conference on Computer Vision (ECCV), 2018, pp. 67–82. 5

    2018

  23. [31]

    Maintaining Discrimination and Fairness in Class Incremental Learning,

    B. Zhao, X. Xiao, G. Gan, B. Zhang, and S.-T. Xia, “Maintaining Discrimination and Fairness in Class Incremental Learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 13208–13217. 8

    2020

  24. [32]

    Human action recognition and prediction: A survey,

    Y. Kong and Y. Fu, “Human action recognition and prediction: A survey,” Int. J. Comput. Vis., vol. 130, no. 5, pp. 1366–1401, 2022. [Online]. Available: https://doi.org/10.1007/s11263-022-01594-9 1

    2022 doi

  25. [33]

    Activitynet: A large-scale video benchmark for human activity understanding,

    F. C. Heilbron, V. Escorcia, B. Ghanem, and J. C. Niebles, “Activitynet: A large-scale video benchmark for human activity understanding,” in IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2015, Boston, MA, USA, June 7-12, 2015 . IEEE Computer Society, 2015, p...

    2015

  26. [34]

    The "something something

    R. Goyal, S. E. Kahou, V. Michalski, J. Materzynska, S. Westphal, H. Kim, V. Haenel, I. Fründ, P. Yianilos, M. Mueller-Freitag, F. Hoppe, C. Thurau, I. Bax, and R. Memisevic, “The "something something" video database for learning and evaluating visual common sense,” inIEEE Int...

    2017 doi

  27. [35]

    UCF101: A dataset of 101 human actions classes from videos in the wild,

    K. Soomro, A. R. Zamir, and M. Shah, “UCF101: A dataset of 101 human actions classes from videos in the wild,”CoRR, vol. abs/1212.0402, 2012. [Online]. Available: http://arxiv.org/abs/1212.0402 1 18 Yuanda Hu et al

    2012 arXiv

  28. [36]

    Interact before align: Leveraging cross-modal knowledge for domain adaptive action recognition,

    L. Yang, Y. Huang, Y. Sugano, and Y. Sato, “Interact before align: Leveraging cross-modal knowledge for domain adaptive action recognition,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022. IEEE, 2022, pp. 14...

    2022

  29. [37]

    Aligning correlation information for domain adaptation in action recognition,

    Y. Xu, H. Cao, K. Mao, Z. Chen, L. Xie, and J. Yang, “Aligning correlation information for domain adaptation in action recognition,” IEEE Trans. Neural Networks Learn. Syst., vol. 35, no. 5, pp. 6767–6778, 2024. [Online]. Available: https://doi.org/10.1109/TNNLS.2022.3212909 2

    2024

  30. [38]

    Video unsupervised domain adaptation with deep learning: A comprehensive survey,

    Y. Xu, H. Cao, Z. Chen, X. Li, L. Xie, and J. Yang, “Video unsupervised domain adaptation with deep learning: A comprehensive survey,” CoRR, vol. abs/2211.10412, 2022. [Online]. Available: https://doi.org/10.48550/arXiv.2211. 10412 2

    2022 doi arXiv

  31. [39]

    ARID: A new dataset for recognizing action in the dark,

    Y. Xu, J. Yang, H. Cao, K. Mao, J. Yin, and S. See, “ARID: A new dataset for recognizing action in the dark,” CoRR, vol. abs/2006.03876, 2020. [Online]. Available: https://arxiv.org/abs/2006.03876 2

    2006 arXiv

  32. [40]

    Continual learning for robotics: Definition, framework, learning strategies, opportunities and challenges,

    T. Lesort, V. Lomonaco, A. Stoian, D. Maltoni, D. Filliat, and N. D. Rodríguez, “Continual learning for robotics: Definition, framework, learning strategies, opportunities and challenges,” Inf. Fusion, vol. 58, pp. 52–68, 2020. [Online]. Available: https://doi.org/10.1016/j.in...

    2020 doi

  33. [41]

    Using noise to compute error surfaces in connectionist networks: A novel means of reducing catastrophic forgetting,

    R. M. French and N. Chater, “Using noise to compute error surfaces in connectionist networks: A novel means of reducing catastrophic forgetting,” Neural Comput., vol. 14, no. 7, pp. 1755–1769, 2002. [Online]. Available: https://doi.org/10.1162/08997660260028700 2

    2002 doi

  34. [42]

    Classifier adaptation at prediction time,

    A. Royer and C. H. Lampert, “Classifier adaptation at prediction time,” inIEEE Conference on Computer Vision and Pattern Recognition, CVPR 2015, Boston, MA, USA, June 7-12, 2015. IEEE Computer Society, 2015, pp. 1401–1409. [Online]. Available: https://doi.org/10.1109/CVPR.2015...

    2015

  35. [43]

    A theory of learning from different domains,

    S. Ben-David, J. Blitzer, K. Crammer, A. Kulesza, F. Pereira, and J. W. Vaughan, “A theory of learning from different domains,”Mach. Learn., vol. 79, no. 1-2, pp. 151–175, 2010. [Online]. Available: https://doi.org/10.1007/s10994-009-5152-4 3

    2010 doi

  36. [44]

    On the stability-plasticity dilemma of class-incremental learning,

    D. Kim and B. Han, “On the stability-plasticity dilemma of class-incremental learning,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17-24, 2023. IEEE, 2023, pp. 20196– 20204. [Online]. Available: https://doi.org/10.1...

    2023

  37. [45]

    Delving into the continuous domain adaptation,

    Y. Xu, Z. Jiang, A. Men, Y. Liu, and Q. Chen, “Delving into the continuous domain adaptation,” in MM ’22: The 30th ACM International Conference on Multimedia, Lisboa, Portugal, October 10 - 14, 2022, J. Magalhães, A. D. Bimbo, S. Satoh, N. Sebe, X. Alameda-Pineda, Q. Jin, V. O...

    2022

  38. [46]

    NTU RGB+D: A large scale dataset for 3d human activity analysis,

    A. Shahroudy, J. Liu, T. Ng, and G. Wang, “NTU RGB+D: A large scale dataset for 3d human activity analysis,” in2016 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2016, Las Vegas, NV, USA, June 27-30, 2016. IEEE Computer Society, 2016, pp. 1010–1019. [Online]...

    2016 doi

  39. [47]

    Etri-activity3d: A large-scale RGB-D dataset for robots to recognize daily activities of the elderly,

    J. Jang, D. Kim, C. Park, M. Jang, J. Lee, and J. Kim, “Etri-activity3d: A large-scale RGB-D dataset for robots to recognize daily activities of the elderly,” in IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2020, Las Vegas, NV, USA, October 24, 202...

    2020

  40. [48]

    Temporal segment networks: Towards good practices for deep action recognition,

    L. Wang, Y. Xiong, Z. Wang, Y. Qiao, D. Lin, X. Tang, and L. V. Gool, “Temporal segment networks: Towards good practices for deep action recognition,” in Computer Vision - ECCV 2016 - 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part ...

    2016 doi

  41. [49]

    An overview of gradient descent optimization algorithms,

    S. Ruder, “An overview of gradient descent optimization algorithms,”CoRR, vol. abs/1609.04747, 2016. [Online]. Available: http://arxiv.org/abs/1609.04747 10

    2016 arXiv

  42. [50]

    Maximum entropy sampling,

    M. C. Shewry and H. P. Wynn, “Maximum entropy sampling,”Journal of applied statistics, vol. 14, no. 2, pp. 165–170, 1987. 14

    1987

  43. [5542]

    4, 5, 8, 10, 12, 13, 14 Abbreviated paper title 17

Pith tools

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