Pith. sign in

REVIEW 2 major objections 5 minor 2 cited by

STNMamba: Mamba-based Spatial-Temporal Normality Learning for Video Anomaly Detection

T0 review · 2 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Mamba model tops video anomaly detection at 7.2M params

desk verdict The efficiency story is real and the Mamba-VAD architecture is worth a look, but the reported SOTA numbers are compromised by test-set hyperparameter selection. read the letter →

arxiv 2412.20084 v1 pith:RBNF6BN5 submitted 2024-12-28 cs.CV

Zhangxun Li , Mengyang Zhao , Xuan Yang , Yang Liu , Jiamu Sheng , Xinhua Zeng , Tian Wang , Kewei Wu
show 1 more author
Yu-Gang Jiang
This is my paper · ORCID
classification cs.CV
keywords videoanomalydetectionMambastatespacemodelsnormalitylearningspatial-temporalconsistencymemorynetworkefficientsurveillancefutureframeprediction
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

STNMamba is a new network for unsupervised video anomaly detection built on Mamba, the state-space sequence model with linear-time complexity. The paper claims that by adapting Mamba's selective scanning to both appearance and motion streams—multi-scale in space, channel-aware in time—and by fusing the two streams at multiple levels with memory-stored normal prototypes, a lightweight model can outperform CNN- and transformer-based detectors. On three benchmarks the method reports frame-level AUCs of 98.0%, 89.0%, and 74.9%, with only 7.2 million parameters and 1.5 GFLOPs per frame. The contribution matters because existing high-accuracy video anomaly detectors are computationally heavy, and the paper argues its design is the first to bring Mamba's efficiency to this task while also improving accuracy.

What carries the argument

The load-bearing mechanism is the Vision Space State Block (VSSB) with 2D-Selective-Scan (SS2D)—the visual adaptation of Mamba's selective scan that flattens a feature map along four directions and processes it with linear complexity. Around this core the paper wraps three additions: MS-VSSB adds parallel depth-wise convolutions (kernels 1×1, 3×3, 5×5) before the VSSB for multi-scale appearance; CA-VSSB adds a channel-attention branch (average and max pooling) to emphasize motion-relevant channels; and the Spatial-Temporal Interaction Module (STIM) uses an STFB (linear + depth-wise conv + Hadamard interaction + another SS2D + ECA channel selection) to project spatial and temporal features into a shared space at each level, followed by a learnable memory bank with top-k reading and softmax writing to store prototypes of normal patterns. This multi-level fusion plus memory constraint is what the paper argues lets Mamba model spatial-temporal consistency without quadratic attention cost.

What would settle it

Measure STNMamba and the top compared baselines (e.g., MNAD and ASTT) on the same GPU with the same input resolution (256×256) and the same number of input frames, and recompute FLOPs and parameters with a standard profiler; if STNMamba's numbers rise to or above the baselines' when measured uniformly, the central efficiency claim is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that Mamba—a selective state-space model—can be made the backbone of a prediction-based video anomaly detector that learns spatial-temporal normality better and far more cheaply than CNN or transformer alternatives. The authors introduce STNMamba, which uses a spatial encoder of Multi-Scale Vision Space State Blocks (MS-VSSB) to aggregate appearance features at several kernel scales, a temporal encoder of Channel-Aware Vision Space State Blocks (CA-VSSB) to emphasize motion channels from RGB frame differences, and a Spatial-Temporal Interaction Module (STIM) that fuses the two streams at every level using Spatial-Temporal Fusion Blocks (STFB) and reads/writes a memory bank of normal prototypes. Anomalies are scored by combining future-frame prediction error (PSNR) with feature distance to the nearest memory item. According to the paper, this achieves state-of-the-art frame-level AUCs of 98.0% on UCSD Ped2, 89.0% on CUHK Avenue, and 74.9% on ShanghaiTech, while running at 40 FPS with 7.2M parameters and 1.5G FLOPs, and it is the first Mamba-based method reported for video anomaly detection.

Load-bearing premise

The load-bearing premise is that the paper's reported FLOPs, parameter counts, and frames-per-second for all compared methods are measured under comparable conditions, since the efficiency advantage over CNN and transformer baselines rests entirely on those numbers.

Editorial extensions

If this is right

  • If the reported efficiency numbers hold under comparable measurement, Mamba-based detectors can serve as a real-time option for video surveillance on edge devices, where transformer-based models are too heavy.
  • The dual-encoder plus multi-level fusion design gives a template for applying state-space models to other video understanding tasks that need spatial-temporal consistency, such as video prediction and action recognition.
  • The memory-bank mechanism in STIM suggests that normal-prototype constraints can be attached at every level of a Mamba encoder, not just the bottleneck, and that this hierarchical memory is what drives the reported gains.
  • The paper's ablations indicate that each component (multi-scale spatial, channel-aware temporal, fusion block, memory, multi-level fusion) contributes additively, so the architecture can be pruned to fit even tighter compute budgets with a predictable drop in AUC.
  • On ShanghaiTech, the largest and most complex benchmark, the reported 74.9% AUC would make Mamba the best prediction-based unsupervised result, which is the strongest evidence for the method's scalability.

Reading between the lines

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

  • The efficiency comparison in Table I mixes numbers the authors reproduced with numbers taken from original papers; if those were measured at different input resolutions or with different frame counts, the 1.5G FLOPs/7.2M params advantage over MNAD and ASTT could shrink, so a standardized measurement protocol would settle this.
  • Mamba's linear complexity should make the gap in FLOPs grow as input resolution increases; a natural test is to measure STNMamba at 640×360 or 856×480 (the native dataset resolutions) rather than the reported 256×256.
  • The normality-score curve in Fig. 6 suggests detection latency is short (scores drop sharply at anomaly onset); a frame-level latency analysis over many videos would quantify whether this holds for rare, gradual anomalies such as loitering.
  • Because the temporal encoder uses only RGB frame differences instead of optical flow, the method may be less sensitive to small, slow motions; evaluating on scenes with subtle anomalies (e.g., a purse snatch) would test whether channel attention compensates for the lack of flow.
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

2 major / 5 minor

Summary. The manuscript proposes STNMamba, a Mamba-based network for unsupervised video anomaly detection. It uses a dual-encoder design (MS-VSSB for multi-scale spatial features, CA-VSSB for temporal motion cues), a Spatial-Temporal Interaction Module with fusion blocks and memory banks, and a decoder that predicts the next frame. Anomaly scores combine prediction PSNR and memory-retrieval distance. On UCSD Ped2, CUHK Avenue, and ShanghaiTech the authors report frame-level AUCs of 98.0%, 89.0%, and 74.9%, respectively, and claim state-of-the-art performance with 1.5G FLOPs, 7.2M parameters, and 40 FPS.

Significance. The architecture is a reasonable and potentially useful first step toward Mamba-based VAD: the dual-encoder with multi-scale and channel-aware Mamba blocks, the multi-level spatial-temporal interaction, and the memory-augmented score are clearly described, and the component ablations (Tables II-IV) show that each proposed block contributes positively. If the reported results are reproducible under a clean evaluation protocol, the efficiency advantage would be practically relevant for edge deployment. The paper is well written and the method is novel relative to the cited CNN/transformer baselines. However, the state-of-the-art claim is currently not supported because the key hyperparameters are tuned on the test sets and the efficiency table is not measured under a single comparable protocol.

major comments (2)
  1. [Section IV-C4, Fig. 5, Eq. (18)] The hyperparameters tau and k are selected by maximizing frame-level AUC on the test sets of UCSD Ped2 and CUHK Avenue, and the same test sets are then used for the reported 98.0% and 89.0% results; the gains over the second-best methods (0.3 and 0.5 points) are of the same scale as this selection bias, so the state-of-the-art claim is not established. Please either fix tau and k on a held-out validation split, report results without test-set selection, or provide repeated-seed statistics with error bars; as written, the 74.9% on ShanghaiTech is also not demonstrably independent of model choices.
  2. [Section IV-B2, Table I] The efficiency comparison is under-specified. The text says some entries were collected from original papers and others reproduced, but it does not state the input resolution, number of frames, or exact GPU and measurement protocol for the reproduced runs, nor how the 1.5G FLOPs, 7.2M parameters, and 40 FPS for STNMamba were measured. Because the central claim of 'fewer parameters and lower computational costs' rests on this table, please report a uniform measurement setup (same resolution, frame count, hardware, and software versions) and ideally release the evaluation code.
minor comments (5)
  1. [Section III-D2, Eq. (11)] The term F_i_m is used in the definition of the memory-enhanced feature but is never defined; please state explicitly that it is the reshaped memory-reconstructed feature obtained from Eq. (7).
  2. [Section III-F, Eq. (14)] The notation m1_i and m2_i is ambiguous because the subscript i simultaneously indexes the query and the rank of the nearest memory item; please define these quantities explicitly.
  3. [Section III-G, Eq. (16)] The PSNR formula should use the square of the maximum pixel value in the numerator; as written, the expression is dimensionally inconsistent.
  4. [Section IV-B2] The statement that MNAD 'achieves slightly faster inference' is inaccurate given Table I, which reports 65 FPS for MNAD versus 40 FPS for STNMamba; please revise the wording.
  5. [Section IV-B2] The comparison with ASTT cites reference [6], but ASTT is reference [12] in the bibliography; please correct the citation.

Circularity Check

1 steps flagged · score 6.0 of 10

Test-set hyperparameter selection makes the reported SOTA AUCs partially self-confirming: τ and k are chosen by maximizing AUC on the same test videos.

  1. fitted input called prediction [Section IV-C4 (Sensitivity to hyperparameters τ and k), Fig. 5; Section IV-B1 (frame-level AUC); Eq. (18); Section III-D2 Eqs. (8)-(9)]
    "The best AUC performance of 98.0% and 89.0% on the UCSD Ped2 and CUHK Avenue datasets is obtained when τ is set to 0.8. ... we obtain the best performance when k is set to 60, which means that we retrieve the top 60% most relevant items in the memory bank during the reading operation."

    The headline claim (Sec. IV-B1: 'frame-level AUCs of 98.0%, 89.0%, and 74.9%') is an evaluation result, but the protocol makes it partly self-confirming. τ is the trade-off in the final score (Eq. 18), and k controls top-k memory retrieval (Eqs. 8-9). Sec. IV-C4/Fig. 5 chooses τ=0.8 and k=60 as the values giving the best AUC on the same UCSD Ped2 and CUHK Avenue test videos that appear in the reported 98.0% and 89.0%. The reported numbers are thus maxima over hyperparameters evaluated on the test data, so the 'prediction' is partly selected by construction from the data it claims to predict; no held-out validation or repeated-seed statistics are given, and the 74.9% ShanghaiTech figure inherits the concern.

full rationale

The architectural derivation itself is self-contained: MS-VSSB, CA-VSSB, STFB, and the memory bank are defined from their own equations and ablations, not from the target AUC. The self-citations (e.g., [22] for RGB difference and spatial-temporal consistency) are background support, not load-bearing. The main circularity is empirical and located in the evaluation protocol: the final anomaly score depends on τ (Eq. 18) and on the top-k retrieval rate (Eqs. 8-9), and Section IV-C4 selects τ=0.8 and k=60 by maximizing AUC on the same test sets whose scores are then reported as state-of-the-art. That makes the headline 98.0/89.0/74.9 partly a selected maximum over test-set-tuned hyperparameters rather than an independent test-set evaluation of a fixed model. This is a partial circularity (fitted input called prediction), not a full reduction of the method to its inputs. The efficiency comparison in Table I is under-specified regarding measurement protocol, but that is a correctness/comparability risk, not circularity.

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

The central claims rest on empirical and architectural assumptions: the validity of prediction-error-based anomaly scoring, the sufficiency of RGB differences for motion, and the trustworthiness of the VSSB backbone. No fundamentally new entities are introduced, but the hyperparameters tau and k are tuned on the evaluation sets, which weakens the independence of the reported results.

free parameters (4)
  • tau (score weight) = 0.8
    Tuned on test sets of UCSD Ped2 and CUHK Avenue via sensitivity analysis; the best test AUC is reported for this value.
  • k (top-k retrieval ratio) = 60%
    Tuned on test sets of UCSD Ped2 and CUHK Avenue; best test AUC at k=60.
  • lambda1, lambda2 (loss weights) = 0.1, 0.01
    Set empirically without reported sensitivity analysis; they control the relative influence of memory compactness and sparsity losses.
  • memory bank sizes N1..N4 = 80, 60, 40, 20
    Empirically set as stated in Section IV-A3; no ablation is provided to justify these values.
assumptions (4)
  • domain assumption Prediction error and memory distance are sufficient to discriminate anomalies from normal events in the frame-level AUC setting.
    The method relies on the standard VAD hypothesis that models trained on normal data produce larger errors for abnormal frames, introduced in Sections I and II.
  • domain assumption RGB frame differences provide a valid replacement for optical flow to capture motion information.
    Stated in Section III-A: 'we simply compute the difference of raw RGB values between adjacent video frames... As demonstrated in previous works'. This avoids optical flow but assumes the motion signal is adequate.
  • standard math The VSSB/SS2D selective scan operation from VMamba behaves as described and can be composed with the proposed multi-scale and channel-attention modifications.
    The method adopts VSSB from VMamba [19] without formal proof; the paper relies on the correctness and efficacy of this prior block (Section III-B1).
  • domain assumption Min-max normalization of anomaly scores is applied per test video and does not distort the AUC comparison.
    Equation 18 uses min-max normalization over the test sequence, following prior work [5]. This is standard practice in VAD evaluations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of STNMamba: Mamba-based Spatial-Temporal Normality Learning for Video Anomaly Detection." pith.science (2026). https://pith.science/paper/RBNF6BN5

@misc{pith2026241220084,
  author       = {Pith},
  title        = {Pith review of: STNMamba: Mamba-based Spatial-Temporal Normality Learning for Video Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RBNF6BN5}},
  note         = {Machine review of arXiv:2412.20084}
}
read the original abstract

Video anomaly detection (VAD) has been extensively researched due to its potential for intelligent video systems. However, most existing methods based on CNNs and transformers still suffer from substantial computational burdens and have room for improvement in learning spatial-temporal normality. Recently, Mamba has shown great potential for modeling long-range dependencies with linear complexity, providing an effective solution to the above dilemma. To this end, we propose a lightweight and effective Mamba-based network named STNMamba, which incorporates carefully designed Mamba modules to enhance the learning of spatial-temporal normality. Firstly, we develop a dual-encoder architecture, where the spatial encoder equipped with Multi-Scale Vision Space State Blocks (MS-VSSB) extracts multi-scale appearance features, and the temporal encoder employs Channel-Aware Vision Space State Blocks (CA-VSSB) to capture significant motion patterns. Secondly, a Spatial-Temporal Interaction Module (STIM) is introduced to integrate spatial and temporal information across multiple levels, enabling effective modeling of intrinsic spatial-temporal consistency. Within this module, the Spatial-Temporal Fusion Block (STFB) is proposed to fuse the spatial and temporal features into a unified feature space, and the memory bank is utilized to store spatial-temporal prototypes of normal patterns, restricting the model's ability to represent anomalies. Extensive experiments on three benchmark datasets demonstrate that our STNMamba achieves competitive performance with fewer parameters and lower computational costs than existing methods.

Figures

Figures reproduced from arXiv: 2412.20084 by the authors.

Figure 1
Figure 1. Performance comparisons with respect to FLOPs and Par [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the main architectures in unsupervi [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of the proposed STNMamba in (a). The structu [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Structure of the proposed Spatial-Temporal Fusion B [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Results of sensitivity analysis to hyperparameters [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: The normality score curves of some test video clips on [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Visualization of predicted frames and their corresp [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. FuseMamba-VD: Dual Branch VideoMamba with Gated Class Token Fusion for Violence Detection

    cs.CV 2025-05 conditional novelty 6.0 of 10

    A dual-branch VideoMamba with gated class-token fusion achieves 95.85% accuracy on a newly combined violence-detection benchmark and 74.13% on DVD, with about half the parameters and FLOPs of the CUE-Net baseline.

  2. Privacy-Preserving Video Anomaly Detection: A Survey

    cs.CV 2024-11 conditional novelty 4.0 of 10

    A new survey organizes privacy-preserving video anomaly detection into a three-branch taxonomy (NIE, DIM, ECI) and catalogs datasets, metrics, and future directions.

Reference graph

Works this paper leans on

40 extracted references · 37 canonical work pages · cited by 2 Pith papers

  1. [1]

    Generalized video anomaly event detection: Sy stematic taxonomy and comparison of deep models,

    Y . Liu, D. Y ang, Y . Wang, J. Liu, J. Liu, A. Boukerche, P . Su n, and L. Song, “Generalized video anomaly event detection: Sy stematic taxonomy and comparison of deep models,” ACM Comput. Surv., vol. 56, no. 7, pp. 189:1–189:38, 2024

    work page 2024

  2. [2]

    L atent space autoregression for novelty detection,

    D. Abati, A. Porrello, S. Calderara, and R. Cucchiara, “L atent space autoregression for novelty detection,” in in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2019, pp. 481–490

    work page 2019

  3. [3]

    Multi-enc oder towards effective anomaly detection in videos,

    Z. Fang, J. T. Zhou, Y . Xiao, Y . Li, and F. Y ang, “Multi-enc oder towards effective anomaly detection in videos,” IEEE Trans. Multimed. , vol. 23, pp. 4106–4116, 2021

    work page 2021

  4. [4]

    Memorizing normality to detect anomaly: Memory- augmented deep autoencoder for unsupervised anomaly detec tion,

    D. Gong, L. Liu, V . Le, B. Saha, M. R. Mansour, S. V enkatesh , and A. van den Hengel, “Memorizing normality to detect anomaly: Memory- augmented deep autoencoder for unsupervised anomaly detec tion,” in in Proc. IEEE/CVF Int. Conf. Comput. Vis. , 2019, pp. 1705–1714

    work page 2019

  5. [5]

    Future frame predicti on for anomaly detection - A new baseline,

    W. Liu, W. Luo, D. Lian, and S. Gao, “Future frame predicti on for anomaly detection - A new baseline,” in in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2018, pp. 6536–6545

    work page 2018

  6. [6]

    Normality lea rning in multispace for video anomaly detection,

    Y . Zhang, X. Nie, R. He, M. Chen, and Y . Yin, “Normality lea rning in multispace for video anomaly detection,” IEEE Trans. Circuits Syst. Video Technol., vol. 31, no. 9, pp. 3694–3706, 2021

    work page 2021

  7. [7]

    Appearance-m otion memory consistency network for video anomaly detection,

    R. Cai, H. Zhang, W. Liu, S. Gao, and Z. Hao, “Appearance-m otion memory consistency network for video anomaly detection,” i n in Proc. AAAI Conf. Artif. Intell. , 2021, pp. 938–946

    work page 2021

  8. [8]

    Video anomaly detection with spatio-temporal dissociation,

    Y . Chang, Z. Tu, W. Xie, B. Luo, S. Zhang, H. Sui, and J. Y uan , “Video anomaly detection with spatio-temporal dissociation,” Pattern Recognit., vol. 122, p. 108213, 2022

    work page 2022

Show all 40 references
  1. [9]

    Memory-enha nced appearance-motion consistency framework for video anomal y detection,

    Z. Ning, Z. Wang, Y . Liu, J. Liu, and L. Song, “Memory-enha nced appearance-motion consistency framework for video anomal y detection,” Comput. Commun. , vol. 216, pp. 159–167, 2024

    2024

  2. [10]

    Anomaly detection in video se quence with appearance-motion correspondence,

    T. Nguyen and J. Meunier, “Anomaly detection in video se quence with appearance-motion correspondence,” in Proc. IEEE/CVF Int. Conf. Comput. Vis., 2019, pp. 1273–1283

    2019

  3. [11]

    Spatio-temp oral unity networking for video anomaly detection,

    Y . Li, Y . Cai, J. Liu, S. Lang, and X. Zhang, “Spatio-temp oral unity networking for video anomaly detection,” IEEE Access , vol. 7, pp. 172 425–172 432, 2019

    2019

  4. [12]

    Transf ormer- based spatio-temporal unsupervised traffic anomaly detect ion in aerial videos,

    T. M. Tran, D. C. Bui, T. V . Nguyen, and K. Nguyen, “Transf ormer- based spatio-temporal unsupervised traffic anomaly detect ion in aerial videos,” IEEE Trans. Circuits Syst. Video Technol. , vol. 34, no. 9, pp. 8292–8309, 2024

    2024

  5. [13]

    Learning memory-guided nor mality for anomaly detection,

    H. Park, J. Noh, and B. Ham, “Learning memory-guided nor mality for anomaly detection,” in in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit., 2020, pp. 14 360–14 369

    2020

  6. [14]

    Video event rest oration based on keyframes for video anomaly detection,

    Z. Y ang, J. Liu, Z. Wu, P . Wu, and X. Liu, “Video event rest oration based on keyframes for video anomaly detection,” in in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. , 2023, pp. 14 592–14 601

    2023

  7. [15]

    Transanom aly: Video anomaly detection using video vision transformer,

    H. Y uan, Z. Cai, H. Zhou, Y . Wang, and X. Chen, “Transanom aly: Video anomaly detection using video vision transformer,” IEEE Access, vol. 9, pp. 123 977–123 986, 2021

    2021

  8. [16]

    Featur e reconstruc- tion with disruption for unsupervised video anomaly detect ion,

    C. Tao, C. Wang, S. Lin, S. Cai, D. Li, and J. Qian, “Featur e reconstruc- tion with disruption for unsupervised video anomaly detect ion,” IEEE Trans. Multim., vol. 26, pp. 10 160–10 173, 2024

    2024

  9. [17]

    Mamba: Linear-time sequence modeling with selective state spaces,

    A. Gu and T. Dao, “Mamba: Linear-time sequence modeling with selective state spaces,” CoRR, vol. abs/2312.00752, 2023

    2023 arXiv

  10. [18]

    Vi sion mamba: Efficient visual representation learning with bidir ectional state space model,

    L. Zhu, B. Liao, Q. Zhang, X. Wang, W. Liu, and X. Wang, “Vi sion mamba: Efficient visual representation learning with bidir ectional state space model,” in in Proc. Int. Conf. Mach. Learn. , 2024

    2024

  11. [19]

    Vmamba: Visual state space model,

    Y . Liu, Y . Tian, Y . Zhao, H. Y u, L. Xie, Y . Wang, Q. Y e, and Y . Liu, “Vmamba: Visual state space model,” CoRR, vol. abs/2401.10166, 2024

    2024 arXiv

  12. [20]

    Vm-unet: Vision mamba unet for med ical image segmentation,

    J. Ruan and S. Xiang, “Vm-unet: Vision mamba unet for med ical image segmentation,” CoRR, vol. abs/2402.02491, 2024

    2024 arXiv

  13. [21]

    Mamba ir: A simple baseline for image restoration with state-space mod el,

    H. Guo, J. Li, T. Dai, Z. Ouyang, X. Ren, and S. Xia, “Mamba ir: A simple baseline for image restoration with state-space mod el,” in Proc. Eur . Conf. Comput. Vis., vol. 15076, 2024, pp. 222–241

    2024

  14. [22]

    Memory-aug mented spatial-temporal consistency network for video anomaly de tection,

    Z. Li, M. Zhao, X. Zeng, T. Wang, and C. Pang, “Memory-aug mented spatial-temporal consistency network for video anomaly de tection,” in in Proc. Chinese Conf. Pattern Recognit. Comput. Vis. , vol. 14430, 2023, pp. 95–107

    2023

  15. [23]

    Spatial-temporal cascade a utoencoder for video anomaly detection in crowded scenes,

    N. Li, F. Chang, and C. Liu, “Spatial-temporal cascade a utoencoder for video anomaly detection in crowded scenes,” IEEE Trans. Multim. , vol. 23, pp. 203–215, 2021

    2021

  16. [24]

    Multi-scale spati otemporal feature fusion network for video saliency prediction,

    Y . Zhang, T. Zhang, C. Wu, and R. Tao, “Multi-scale spati otemporal feature fusion network for video saliency prediction,” IEEE Trans. Multim., vol. 26, pp. 4183–4193, 2024

    2024

  17. [25]

    A motion-appeara nce-aware network for object change detection,

    H. Zhang, S. Qu, H. Li, W. Xu, and X. Du, “A motion-appeara nce-aware network for object change detection,” Knowl. Based Syst. , vol. 255, p. 109612, 2022

    2022

  18. [26]

    Efficiently modeling long seq uences with structured state spaces,

    A. Gu, K. Goel, and C. R´ e, “Efficiently modeling long seq uences with structured state spaces,” in in Proc. Int. Conf. Learn. Represent. , 2022

    2022

  19. [27]

    Eca-net: E fficient channel attention for deep convolutional neural networks,

    Q. Wang, B. Wu, P . Zhu, P . Li, W. Zuo, and Q. Hu, “Eca-net: E fficient channel attention for deep convolutional neural networks, ” in in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. , 2020, pp. 11 531– 11 539

    2020

  20. [28]

    Anomaly detec tion and localization in crowded scenes,

    W. Li, V . Mahadevan, and N. V asconcelos, “Anomaly detec tion and localization in crowded scenes,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 36, no. 1, pp. 18–32, 2014

    2014

  21. [29]

    Abnormal event detection at 15 0 FPS in MA TLAB,

    C. Lu, J. Shi, and J. Jia, “Abnormal event detection at 15 0 FPS in MA TLAB,” in in Proc. IEEE Int. Conf. Comput. Vis. , 2013, pp. 2720– 2727

    2013

  22. [30]

    Adam: A method for stochastic opt imization,

    D. P . Kingma and J. Ba, “Adam: A method for stochastic opt imization,” in in Proc. Int. Conf. Learn. Represent. , 2015

    2015

  23. [31]

    Abnormal event detection in videos using gener ative adversarial nets,

    M. Ravanbakhsh, M. Nabi, E. Sangineto, L. Marcenaro, C. S. Regazzoni, and N. Sebe, “Abnormal event detection in videos using gener ative adversarial nets,” in Proc. IEEE Int. Conf. Image Process. , 2017, pp. 1577–1581

    2017

  24. [32]

    Remembering history with con volutional LSTM for anomaly detection,

    W. Luo, W. Liu, and S. Gao, “Remembering history with con volutional LSTM for anomaly detection,” in Proc. IEEE Int. Conf. Multimedia Expo., 2017, pp. 439–444

    2017

  25. [33]

    Video anomaly detection with sparse coding inspired deep neural n etworks,

    W. Luo, W. Liu, D. Lian, J. Tang, L. Duan, X. Peng, and S. Ga o, “Video anomaly detection with sparse coding inspired deep neural n etworks,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 43, no. 3, pp. 1070–1084, 2021

    2021

  26. [34]

    Anopcn: Video a nomaly detection via deep predictive coding network,

    M. Y e, X. Peng, W. Gan, W. Wu, and Y . Qiao, “Anopcn: Video a nomaly detection via deep predictive coding network,” in in Proc. ACM Int. Conf. Multimedia., 2019, pp. 1805–1813

    2019

  27. [35]

    Attentio n-based anomaly detection in multi-view surveillance videos,

    Q. Li, R. Y ang, F. Xiao, B. Bhanu, and F. Zhang, “Attentio n-based anomaly detection in multi-view surveillance videos,” Knowl. Based Syst., vol. 252, p. 109348, 2022

    2022

  28. [36]

    Spatiotempor al consistency-enhanced network for video anomaly detection ,

    Y . Hao, J. Li, N. Wang, X. Wang, and X. Gao, “Spatiotempor al consistency-enhanced network for video anomaly detection ,” Pattern Recognit., vol. 121, p. 108232, 2022

    2022

  29. [37]

    Multi-branch ga n-based abnormal events detection via context learning in surveill ance videos,

    D. Li, X. Nie, R. Gong, X. Lin, and H. Y u, “Multi-branch ga n-based abnormal events detection via context learning in surveill ance videos,” IEEE Trans. Circuits Syst. Video Technol., vol. 34, no. 5, pp. 3439–3450, 2024

    2024

  30. [38]

    Attention-based residual autoencode r for video anomaly detection,

    V . Le and Y . Kim, “Attention-based residual autoencode r for video anomaly detection,” Appl. Intell. , vol. 53, no. 3, pp. 3240–3254, 2023

    2023

  31. [39]

    V ideo anomaly detection guided by clustering learning,

    S. Qiu, J. Y e, J. Zhao, L. He, L. Liu, B. E., and X. Huang, “V ideo anomaly detection guided by clustering learning,” Pattern Recognit., vol. 153, p. 110550, 2024

    2024

  32. [40]

    Long short-term dyn amic prototype alignment learning for video anomaly detection,

    C. Huang, J. Wen, C. Liu, and Y . Liu, “Long short-term dyn amic prototype alignment learning for video anomaly detection, ” in Proc. Thirty-Third Int. Joint Conf. Artif. Intell. , 2024, pp. 866–874

    2024

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.