Pith. sign in

REVIEW 4 major objections 8 minor 1 cited by

A Multimodal Deviation Perceiving Framework for Weakly-Supervised Temporal Forgery Localization

T0 review · 4 major / 8 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A weakly-supervised framework localizes forged video segments with only video-level labels and approaches fully supervised accuracy.

desk verdict Genuinely new WS-TFL framework with helpful ablations, but the deviation loss is an untested assumption and the fully-supervised comparison claim only holds at loose IoU. read the letter →

arxiv 2507.16596 v2 pith:NW4ODDMS submitted 2025-07-22 cs.CV

Wenbo Xu , Junyan Wu , Wei Lu , Xiangyang Luo , Qian Wang This is my paper · ORCID
classification cs.CV
keywords deepfakedetectiontemporalforgerylocalizationweakly-supervisedlearningmultimodalcross-modalattentiondeviationperceivinglossvideoforensicsaudio-visual
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

This paper aims to prove that temporal forgery localization—finding the start and end timestamps of manipulated segments inside a deepfake video—can be done with only video-level fake/real labels, without frame-level timestamp annotations. The proposed MDP framework reads visual and audio features, aligns them in time, fuses them with cross-modal attention that preserves temporal ordering, and adds a deviation-perceiving loss that widens the gap between adjacent segments in forged videos while shrinking it in genuine ones. On the LAV-DF benchmark, MDP reaches 84.57 AP@0.5 and 75.91 AP@0.75 using only weak labels, closing much of the gap to fully supervised methods, and on AV-Deepfake1M it localizes very short forged segments in long videos. A sympathetic reader would take the paper's contribution to be evidence that weak supervision, combined with multimodal deviation, is a viable route to scalable deepfake forensics.

What carries the argument

Two mechanisms carry the argument. The first is the multimodal interaction module, built around a temporal-property-preserving cross-modal attention: visual and audio features are mapped into a probabilistic embedding space by an MLP with LayerNorm, and attention is computed between the modalities in a way that sums the relevance matrix over columns before weighting values, so the temporal index is never collapsed. The second is the extensible deviation-perceiving loss $\mathcal{L}_{dp}$: the temporal deviation $d$ is the normalized sum over adjacent segments of a deviation function $f$ (mean squared error in the main experiments), and the loss is binary cross-entropy between $d$ and the video-level label, enlarging adjacent-segment deviation for forged videos and reducing it for genuine ones. Together they convert weak video-level labels into segment-level supervision for timestamp localization.

What would settle it

Train MDP on a set where genuine videos contain frequent scene cuts and forged segments are produced by a smooth model that preserves temporal coherence across the splice boundary; if the deviation-perceiving loss still separates forged from genuine videos by pushing adjacent-segment deviation apart, the core assumption survives, and if it does not, the loss is actively misleading the model.

Watch

Extended reading notes

Core claim

The central claim is that inter-modality and inter-segment deviations carry enough forgery signal to support weakly-supervised temporal forgery localization. The framework extracts frame-level visual and audio features, tokenizes them, pools both modalities to a common temporal length, then computes cross-modal attention in a probabilistic embedding space: query from one modality, key and value from the other, with the relevance matrix summed over columns so temporal position survives the attention step. The enhanced visual and audio features are concatenated into a comprehensive video feature, from which a temporal forgery activation sequence is produced. Training uses only the video-level label via a classification loss plus the deviation-perceiving loss, which measures the mean squared error between adjacent segment features, normalizes it, and applies binary cross-entropy against the video label. The result is a model that predicts forged segment timestamps at inference and, in several evaluation metrics on LAV-DF and AV-Deepfake1M, performs close to fully supervised temporal forgery localization methods.

Load-bearing premise

The load-bearing premise is that forged splicing makes adjacent video segments measurably less smooth than genuine footage, and that this smoothness gap shows up in the pooled visual and audio features and can be supervised by a single whole-video label.

Editorial extensions

If this is right

  • Temporal forgery localization becomes trainable on large video collections that only carry whole-video labels, removing the main annotation bottleneck of frame-level supervision.
  • The temporal-preserving cross-modal attention is a reusable fusion pattern for any multimodal video task where order in time must survive attention.
  • The deviation-perceiving loss is plug-compatible with different deviation measures; the paper shows L1, L2, and MSE behave differently, so better measures can extend the same framework.
  • The reported AP@0.95 on LAV-DF remains far below fully supervised methods, so the next concrete step is refining boundary precision rather than coarse segment finding.
  • On AV-Deepfake1M, MDP localized forged segments as short as roughly 0.3 seconds inside a 23-second video using only video-level labels, indicating the method scales to long videos.

Reading between the lines

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

  • One consequence the authors leave implicit: the deviation idea is not tied to audio-visual video, and the same loss could be applied to partially manipulated speech, image strips, or any sequence where forgery introduces discontinuity.
  • A direct stress test the paper does not run is genuine footage with scene cuts or rapid camera motion—these would inflate adjacent-segment deviation and could make the loss push genuine videos in the wrong direction.
  • The paper reports results for real and audio-visual forgery samples only; audio-only and visual-only forgeries are untested, so the relative contribution of each modality to deviation remains unknown.
  • Because deviation is computed on concatenated multimodal features, decomposing the loss per modality could reveal which stream carries the forgery trace, a diagnostic extension the framework supports.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 8 minor

Summary. The paper proposes MDP, a weakly-supervised temporal forgery localization (WS-TFL) framework that trains with only video-level genuine/forged labels. It extracts visual and audio features with pretrained backbones, aligns them to T segments, applies a temporal-property-preserving cross-modal attention in a probabilistic embedding space, and produces a temporal forgery activation sequence (FAS) from a classifier (Sections 3.2–3.3). Training combines a video-level classification loss with a deviation perceiving loss (Eqs. 9–12) that enlarges a scalar measure of adjacent-segment feature deviation for forged videos and shrinks it for genuine videos. Experiments on LAV-DF and AV-Deepfake1M report AP and AR; ablations on LAV-DF attribute gains to both the cross-modal attention and the deviation loss, and a comparison of L1/L2/MSE deviation measures is included (Table 4). The abstract and introduction claim results comparable to fully-supervised approaches in several metrics.

Significance. If the empirical results hold, MDP would provide a simple, well-specified baseline for WS-TFL, a task with clear practical relevance; its gains over transferred WS-TAL baselines are large (LAV-DF AP@0.5: 84.57 vs. 18.79/8.79 for FuSTAL/CoLA), both proposed components are ablated, the deviation measure function is explored in Table 4, and the code is released. The significance is tempered, though, because 'comparable to fully-supervised' holds only at lenient IoU on one dataset or at AP@0.1/0.2 on the other, and because the deviation prior—the only temporally non-uniform training signal—is verified only indirectly through ablation rather than by direct measurement.

major comments (4)
  1. [Section 3.4, Eqs. (9)–(11) and Table 3] The deviation perceiving loss is the only training objective with temporal granularity—L_cls receives a uniform gradient over segments through Eq. (1)—yet L_dp supervises a single scalar d aggregated over all T adjacent pairs, so by construction it can only bias the entire activation sequence and cannot point to the correct boundary; the ablation attributes about +11.7 average AP to it. The load-bearing premise that forged videos have larger adjacent-segment deviation than genuine videos is not directly verified on LAV-DF or AV-Deepfake1M; the cited refs [18,42] are video-level detectors, and genuine scene cuts or temporally smooth forgeries are natural counterexamples. Please add a diagnostic of the prior (e.g., distribution of d and of per-boundary deviations on held-out sets, overlap of top-deviance boundaries with GT boundaries) and, if possible, an ablation in which L_dp is replaced by per-segment or boundary-level pseudo-supervision.
  2. [Section 3.3, Eqs. (6)–(7) and Algorithm 1] The core cross-modal attention formula is dimensionally inconsistent as written: bR is defined as a 1×T row vector (column sum of R) and V is T×d, so bR^T·V is not a valid matrix product; presumably the intended operation is row-wise scaling of each audio segment by its relevance r_t, i.e., ATT_v[t] = r_t·V[t]. Please rewrite with explicit tensor shapes, and define cM_v and cM_a, which appear in Eq. (8) but are never introduced.
  3. [Sections 3.2 and 3.5] The inference procedure is underspecified: no threshold is given for converting p_t into binary forged/background segment decisions, no rule is given for merging consecutive forged segments, and the AR evaluation (proposals of 20/10/5/2) requires a proposal-generation mechanism that is not described. These omissions block reproduction of the reported AP/AR numbers; please specify the threshold (or the selection procedure) and the proposal generation/merging algorithm.
  4. [Abstract and Section 4.2, Table 2] The claim of 'comparable results to fully-supervised approaches' is only supportable at lenient IoU thresholds on LAV-DF (AP@0.5, AP@0.75) and at AP@0.1/AP@0.2 on AV-Deepfake1M; on AV-Deepfake1M the average AR is 8.86 versus 87.88–88.62 for fully-supervised methods and AP@0.5 drops to 22.39 versus 99.25+ for fully-supervised methods. Please temper the abstract and contributions accordingly and discuss the AR collapse on long-duration videos, since this is the localization metric where the weak supervision is most strained.
minor comments (8)
  1. [Section 3.5, Eq. (12)] L_cls is referenced but never defined; please give the explicit loss form (presumably cross-entropy between ŷ and y).
  2. [Section 4.1] The segment count T and the choice of φ=0.5 are not justified, and no sensitivity analysis over φ is reported; please add one or state why these values are sufficient.
  3. [Section 3.3] The text says the attention measures relevance 'between each video segment v_t and audio segment a_t' but then defines a full T×T relevance matrix; please reconcile the wording.
  4. [Eq. (9)] The sum over t=1..T includes a boundary pair (x_T, x_{T+1}); please state how the last segment is handled.
  5. [Figure 2 and inline math] Several passages contain garbled or duplicated tokens (e.g., 'VVisual features', repeated labels next to arrows, symbol fragments such as 'dp', 'cls'); please check the camera-ready rendering.
  6. [Section 4.1] It is stated that only real and audio-visual forgery samples are used; please report the number of videos retained per dataset and how the forging masks are obtained, since this affects comparability with the fully-supervised baselines.
  7. [Table 2] The absence of CoLA and FuSTAL rows is explained only in the body text; please add a table note to make this visible.
  8. [Experimental protocol] No error bars or multiple-seed results are reported; please state the number of runs and any variance, especially for the ablation deltas in Table 3.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the weakly-supervised losses are standard training signals and the localization output is not used as a training target.

full rationale

The derivation chain is self-contained. The framework's only training signals are the video-level classification loss (Eq. 1) and the deviation perceiving loss (Eqs. 9-11), both supervised by the video-level label y_i. The inference output—start/end timestamps derived from the FAS P—is never used as a training target, so there is no fitted-input-called-prediction loop. L_dp does not define the FAS in terms of the predicted timestamps; it is a regularizer that encodes a distributional prior (forged videos have larger adjacent-segment deviation) and its contribution is measured by an ablation (Table 3), which is an external, falsifiable check rather than a circular reduction. The prior is cited to [18,42], which include co-author Wei Lu, but those are published external results and the paper supplements them with its own rationale and experiments; this is not a load-bearing self-citation chain of the kind that would force the conclusion. The multimodal interaction (Eqs. 6-8) is a standard attention construction, not a renaming of the target. Overall, no step in the claimed derivation reduces by construction to its own inputs.

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

The ledger is light: no new entities, and the main load-bearing assumptions are the temporal deviation premises of the loss, which are supported only by the authors' own prior work. The free parameters are mostly standard hyperparameters, but the inference threshold being unspecified is a gap.

free parameters (4)
  • phi (loss balance weight) = 0.5
    Balance weight for deviation loss, set by hand.
  • T (temporal segment count) = not reported
    Number of temporal segments after pooling; a design choice that affects evaluation.
  • Inference threshold for segment classification = not specified
    Threshold for labeling a segment as forged during inference; unspecified in Section 3.5.
  • Deviation measure function f = MSE
    Selected among L1, L2, MSE based on LAV-DF performance (Table 4), so effectively a fitted choice.
assumptions (4)
  • domain assumption Adjacent frames in genuine videos exhibit minimal changes in content and statistical properties.
    Stated in Section 3.4 as the basis for the deviation loss.
  • domain assumption Forged segments have larger deviation from adjacent segments than genuine samples.
    Central premise of L_dp, Section 3.4, supported by citations [18, 42] which include co-author Wei Lu.
  • domain assumption Pre-trained feature extractors (TSN, Wav2Vec, ResNet) provide fixed informative features.
    Used in Section 4.1; standard practice, but the choice of extractor affects results.
  • domain assumption Probabilistic embedding space improves cross-modal representation.
    Borrowed from [16] (Probabilistic Vision-Language Representation), used in Eq. 4-5.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Multimodal Deviation Perceiving Framework for Weakly-Supervised Temporal Forgery Localization." pith.science (2026). https://pith.science/paper/NW4ODDMS

@misc{pith2026250716596,
  author       = {Pith},
  title        = {Pith review of: A Multimodal Deviation Perceiving Framework for Weakly-Supervised Temporal Forgery Localization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NW4ODDMS}},
  note         = {Machine review of arXiv:2507.16596}
}
read the original abstract

Current researches on Deepfake forensics often treat detection as a classification task or temporal forgery localization problem, which are usually restrictive, time-consuming, and challenging to scale for large datasets. To resolve these issues, we present a multimodal deviation perceiving framework for weakly-supervised temporal forgery localization (MDP), which aims to identify temporal partial forged segments using only video-level annotations. The MDP proposes a novel multimodal interaction mechanism (MI) and an extensible deviation perceiving loss to perceive multimodal deviation, which achieves the refined start and end timestamps localization of forged segments. Specifically, MI introduces a temporal property preserving cross-modal attention to measure the relevance between the visual and audio modalities in the probabilistic embedding space. It could identify the inter-modality deviation and construct comprehensive video features for temporal forgery localization. To explore further temporal deviation for weakly-supervised learning, an extensible deviation perceiving loss has been proposed, aiming at enlarging the deviation of adjacent segments of the forged samples and reducing that of genuine samples. Extensive experiments demonstrate the effectiveness of the proposed framework and achieve comparable results to fully-supervised approaches in several evaluation metrics.

Figures

Figures reproduced from arXiv: 2507.16596 by the authors.

Figure 1
Figure 1. The schematic diagram of weakly-supervised tem [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Diagrammatic overview of the proposed multimodal deviation perceiving framework for weakly-supervised temporal [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visualization results on the LAV-DF and AV-Deepfake1M. For LAV-DF, two fully-supervised approaches UMMAFormer, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 3
Figure 3. Figure 3: From the visualization results, it could be observed that MDP could localize the timestamps of all the forged segments more pre￾cisely compared to the CoLA and FuSTAL on LAV-DF. Moreover, the localization effectiveness of MDP is comparable to that observed in fully-sup…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Registering the 4D Millimeter Wave Radar Point Clouds Via Generalized Method of Moments

    cs.RO 2025-08 unverdicted novelty 4.0 of 10

    The abstract claims a correspondence-free 4D radar registration method based on the Generalized Method of Moments, but the submitted full text is an unrelated Deepfake detection preprint.

Reference graph

Works this paper leans on

45 extracted references · 41 canonical work pages · cited by 1 Pith paper

  1. [1]

    Alexei Baevski, Yuhao Zhou, Abdelrahman Mohamed, and Michael Auli. 2020. wav2vec 2.0: A Framework for Self-supervised Learning of Speech Representa- tions. Advances in neural information processing systems 33 (2020), 12449–12460

    work page 2020

  2. [2]

    Zhixi Cai, Shreya Ghosh, Aman Pankaj Adatia, Munawar Hayat, Abhinav Dhall, Tom Gedeon, and Kalin Stefanov. 2024. AV-Deepfake1M: A Large-Scale LLM- Driven Audio-Visual Deepfake Dataset. In Proceedings of the 32nd ACM Interna- tional Conference on Multimedia . 7414–7423

    work page 2024

  3. [3]

    Zhixi Cai, Shreya Ghosh, Abhinav Dhall, Tom Gedeon, Kalin Stefanov, and Munawar Hayat. 2023. Glitch in the matrix: A large scale benchmark for content driven audio–visual forgery detection and localization. Computer Vision and Image Understanding 236 (2023), 103818

    work page 2023

  4. [4]

    Zhixi Cai, Kalin Stefanov, Abhinav Dhall, and Munawar Hayat. 2022. Do You Re- ally Mean That? Content Driven Audio-Visual Deepfake Dataset and Multimodal Method for Temporal Forgery Localization. In 2022 International Conference on Digital Image Computing: Techniques and Applications . 1–10

    work page 2022

  5. [5]

    Guan-Lin Chen and Chih-Chung Hsu. 2023. Jointly Defending DeepFake Manip- ulation and Adversarial Attack Using Decoy Mechanism. IEEE Transactions on Pattern Analysis and Machine Intelligence 45, 8 (2023), 9922–9931

    work page 2023

  6. [6]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A Simple Framework for Contrastive Learning of Visual Representations. In International Conference on Machine Learning . 1597–1607

    work page 2020

  7. [7]

    Zhiwei Chen, Siwei Wang, Liujuan Cao, Yunhang Shen, and Rongrong Ji. 2024. Adaptive Zone Learning for Weakly Supervised Object Localization. IEEE Trans- actions on Neural Networks and Learning Systems (2024)

    work page 2024

  8. [8]

    Komal Chugh, Parul Gupta, Abhinav Dhall, and Ramanathan Subramanian. 2020. Not made for each other- Audio-Visual Dissonance-based Deepfake Detection and Localization. In Proceedings of the 28th ACM International Conference on Multimedia. 439–447

    work page 2020

Show all 45 references
  1. [9]

    Shichao Dong, Jin Wang, Renhe Ji, Jiajun Liang, Haoqiang Fan, and Zheng Ge

  2. [10]

    Qianhan Feng, Wenshuo Li, Tong Lin, and Xinghao Chen. 2024. Full-Stage Pseudo Label Quality Enhancement for Weakly-supervised Temporal Action Localization. arXiv preprint arXiv:2407.08971 (2024)

    2024 arXiv

  3. [11]

    Haolong Fu, Jin Yuan, Guojin Zhong, Xuan He, Jiacheng Lin, and Zhiyong Li

  4. [12]

    Zhihao Gu, Taiping Yao, Yang Chen, Ran Yi, Shouhong Ding, and Lizhuang Ma. 2022. Region-Aware Temporal Inconsistency Learning for DeepFake Video Detection. 920–926

    2022

  5. [13]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep Residual Learning for Image Recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 770–778

    2016

  6. [14]

    Shan Jia, Reilin Lyu, Kangran Zhao, Yize Chen, Zhiyuan Yan, Yan Ju, Chuanbo Hu, Xin Li, Baoyuan Wu, and Siwei Lyu. 2024. Can ChatGPT Detect Deepfakes? A Study of Using Multimodal Large Language Models for Media Forensics. In Proceedings of the IEEE/CVF Conference on Computer V...

    2024

  7. [15]

    Zecheng Li, Zening Zeng, Yuqi Liang, and Jin-Gang Yu. 2024. Complete In- stances Mining for Weakly Supervised Instance Segmentation. arXiv preprint arXiv:2402.07633 (2024)

    2024 arXiv

  8. [16]

    Geuntaek Lim, Hyunwoo Kim, Joonsoo Kim, and Yukyung Choi. 2024. Proba- bilistic Vision-Language Representation for Weakly Supervised Temporal Action Localization. In Proceedings of the 32nd ACM International Conference on Multi- media. 5507–5516

    2024

  9. [17]

    Miao Liu, Jing Wang, Xinyuan Qian, and Haizhou Li. 2024. Audio-Visual Tempo- ral Forgery Detection Using Embedding-Level Fusion and Multi-Dimensional Contrastive Loss. IEEE Transactions on Circuits and Systems for Video Technology 34, 8 (2024), 6937–6948

    2024

  10. [18]

    Wei Lu, Lingyi Liu, Bolin Zhang, Junwei Luo, Xianfeng Zhao, Yicong Zhou, and Jiwu Huang. 2024. Detection of Deepfake Videos Using Long-Distance Attention. IEEE Transactions on Neural Networks and Learning Systems 35, 7 (2024), 9366–9379

    2024

  11. [19]

    Harry McGurk and John MacDonald. 1976. Hearing lips and seeing voices.Nature 264, 5588 (1976), 746–748

    1976

  12. [20]

    Fan Nie, Jiangqun Ni, Jian Zhang, Bin Zhang, and Weizhe Zhang. 2024. FRADE: Forgery-aware Audio-distilled Multimodal Learning for Deepfake Detection. In Proceedings of the 32nd ACM International Conference on Multimedia . 6297–6306

    2024

  13. [21]

    Daisuke Niizumi, Daiki Takeuchi, Yasunori Ohishi, Noboru Harada, and Kunio Kashino. 2021. BYOL for Audio: Self-Supervised Learning for General-Purpose Audio Representation. In 2021 International Joint Conference on Neural Networks . 1–8

    2021

  14. [22]

    Yuval Nirkin, Lior Wolf, Yosi Keller, and Tal Hassner. 2021. DeepFake Detection Based on Discrepancies Between Faces and Their Context. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 10 (2021), 6111–6121

    2021

  15. [23]

    Ziqi Sheng, Wei Lu, Xiangyang Luo, Jiantao Zhou, and Xiaochun Cao. 2025. SUMI- IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 720–728

    2025

  16. [24]

    Ziqi Sheng, Zuomin Qu, Wei Lu, Xiaochun Cao, and Jiwu Huang. 2024. DiRLoc: Disentanglement Representation Learning for Robust Image Forgery Localization. IEEE Transactions on Dependable and Secure Computing (2024)

    2024

  17. [25]

    Dingfeng Shi, Yujie Zhong, Qiong Cao, Lin Ma, Jia Li, and Dacheng Tao. 2023. TriDet: Temporal Action Detection With Relative Boundary Modeling. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 18857–18866

    2023

  18. [26]

    Nina Shvetsova, Brian Chen, Andrew Rouditchenko, Samuel Thomas, Brian Kingsbury, Rogerio S Feris, David Harwath, James Glass, and Hilde Kuehne. 2022. Everything at Once-Multi-modal Fusion Transformer for Video Retrieval. In Proceedings of the IEEE/CVF Conference on Computer Vi...

    2022

  19. [27]

    Hao Su and Meng Yang. 2024. A Consistency and Integration Model with Adaptive Thresholds for Weakly Supervised Object Localization. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence . 1281–1289

    2024

  20. [28]

    Binglu Wang, Yongqiang Zhao, Le Yang, Teng Long, and Xuelong Li. 2023. Tem- poral Action Localization in the Deep Learning Era: A Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence (2023)

    2023

  21. [29]

    Limin Wang, Yuanjun Xiong, Zhe Wang, Yu Qiao, Dahua Lin, Xiaoou Tang, and Luc Van Gool. 2018. Temporal Segment Networks for Action Recognition in Videos. IEEE Transactions on Pattern Analysis and Machine Intelligence 41, 11 (2018), 2740–2755

    2018

  22. [30]

    Junyan Wu, Wenbo Xu, Wei Lu, Xiangyang Luo, Rui Yang, and Shize Guo. 2025. Weakly-supervised Audio Temporal Forgery Localization via Progressive Audio- language Co-learning Network. arXiv preprint arXiv:2505.01880 (2025)

    2025 arXiv

  23. [31]

    Ruiyang Xia, Decheng Liu, Jie Li, Lin Yuan, Nannan Wang, and Xinbo Gao. 2024. MMNet: Multi-Collaboration and Multi-Supervision Network for Sequential Deepfake Detection. IEEE Transactions on Information Forensics and Security 19 (2024), 3409–3422

    2024

  24. [32]

    Qilin Yin, Wei Lu, Xiaochun Cao, Xiangyang Luo, Yicong Zhou, and Jiwu Huang

  25. [33]

    Qilin Yin, Wei Lu, Bin Li, and Jiwu Huang. 2023. Dynamic Difference Learn- ing With Spatio–Temporal Correlation for Deepfake Video Detection. IEEE Transactions on Information Forensics and Security (2023)

    2023

  26. [34]

    Wulian Yun, Mengshi Qi, Chuanming Wang, and Huadong Ma. 2024. Weakly- Supervised Temporal Action Localization by Inferring Salient Snippet-Feature. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 38. 6908–6916

    2024

  27. [35]

    International Journal of Computer Vision (2024), 1–15

    Fine-Grained Multimodal DeepFake Classification via Heterogeneous Graphs. International Journal of Computer Vision (2024), 1–15

    2024

  28. [36]

    Chen-Lin Zhang, Jianxin Wu, and Yin Li. 2022. ActionFormer: Localizing Mo- ments of Actions with Transformers. In European Conference on Computer Vision . 492–510

    2022

  29. [37]

    Dingwen Zhang, Junwei Han, Gong Cheng, and Ming-Hsuan Yang. 2021. Weakly Supervised Object Localization and Detection: A Survey. IEEE Transactions on Pattern Analysis and Machine Intelligence 44, 9 (2021), 5866–5885

    2021

  30. [38]

    Can Zhang, Meng Cao, Dongming Yang, Jie Chen, and Yuexian Zou. 2021. Cola: Weakly-supervised temporal action localization with snippet contrastive learn- ing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 16010–16019

    2021

  31. [39]

    Yi Zhang, Changtao Miao, Man Luo, Jianshu Li, Wenzhong Deng, Weibin Yao, Zhe Li, Bingyu Hu, Weiwei Feng, Tao Gong, et al. 2024. MFMS: Learning Modality- Fused and Modality-Specific Features for Deepfake Detection and Localization Tasks. In Proceedings of the 32nd ACM Internati...

    2024

  32. [40]

    Yipin Zhou and Ser-Nam Lim. 2021. Joint Audio-Visual Deepfake Detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 14800– 14809

    2021

  33. [41]

    Rui Zhang, Hongxia Wang, Mingshan Du, Hanqing Liu, Yang Zhou, and Qiang Zeng. 2023. UMMAFormer: A Universal Multimodal-adaptive Transformer Frame- work for Temporal Forgery Localization. In Proceedings of the 31st ACM Interna- tional Conference on Multimedia . 8749–8759

    2023

  34. [42]

    Chuntao Zhu, Bolin Zhang, Qilin Yin, Chengxi Yin, and Wei Lu. 2024. Deepfake detection via inter-frame inconsistency recomposition and enhancement. Pattern Recognition 147 (2024), 110077

    2024

  35. [44]

    Zhi-Hua Zhou. 2018. A brief introduction to weakly supervised learning.National science review 5, 1 (2018), 44–53

    2018

  36. [2023]

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

    Implicit Identity Leakage: The Stumbling Block to Improving Deepfake Detection Generalization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 3994–4004

  37. [2024]

    In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence

    CF-Deformable DETR: An End-to-End Alignment-Free Model for Weakly Aligned Visible-Infrared Object Detection. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence . 758–766

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.