Research map / Video / Video segmentation and tracking
Deepfake video detection
Detects forged face videos via inconsistency learning, facial components and self-supervised real-face cues.
47 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Video segmentation and tracking, keyword cluster “detection · anomaly · supervised”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 4 tracked benchmarks.
13 papers in 2023–24 and 15 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 3.0% to 2.8%.
Open questions the data points to
- DFEW (WAR): best went from 69.25 (up to 2024) to 77.45 in 2025 — AVF-MAE++: Scaling Affective Video Facial Masked Autoencoders via Efficient Audio-Visual Self-Supervised Learning (CVPR 2025 · 7 citations)
No longer reported by papers here: Celeb-DF (AUC (%), last in 2023); DFDC (AUC (%), last in 2023); FF++ (AUC, last in 2024).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · LOMo: Latent Ordinal Model for Facial Analysis in Videos (CVPR 2016 · 97 citations)
- 2017 · Neural Aggregation Network for Video Face Recognition (CVPR 2017 · 420 citations)
- 2018 · Deception Detection in Videos (AAAI 2018 · 140 citations)
- 2019 · Face-Focused Cross-Stream Network for Deception Detection in Videos (CVPR 2019 · 76 citations)
- 2020 · Celeb-DF: A Large-Scale Challenging Dataset for DeepFake Forensics (CVPR 2020 · 1,725 citations)
- 2021 · Lips Don't Lie: A Generalisable and Robust Approach To Face Forgery Detection (CVPR 2021 · 501 citations)
- 2022 · Leveraging Real Talking Faces via Self-Supervision for Robust Forgery Detection (CVPR 2022 · 166 citations)
- 2023 · AltFreezing for More General Video Face Forgery Detection (CVPR 2023 · 126 citations)
- 2024 · Exploiting Style Latent Flows for Generalizing Deepfake Video Detection (CVPR 2024 · 77 citations)
- 2025 · Towards More General Video-based Deepfake Detection through Facial Component Guided Adaptation for Foundation Model (CVPR 2025 · 23 citations)
- 2026 · Tell me Habibi, is it Real or Fake? (ICLR 2026 · 0 citations)
Earliest papers here
- LOMo: Latent Ordinal Model for Facial Analysis in Videos (CVPR 2016 · 97 citations)
- Neural Aggregation Network for Video Face Recognition (CVPR 2017 · 420 citations)
- Joint Registration and Representation Learning for Unconstrained Face Identification (CVPR 2017 · 39 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
Celeb-DF · AUC (%) (higher is better)
4 papers, 2021–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 56.9 | Improving the Efficiency and Robustness of Deepfakes Detection Through Precise Geometric Features (CVPR 2021 · 172 citations) |
| 2022 | 93.3 | FInfer: Frame Inference-Based Deepfake Detection for High-Visual-Quality Videos (AAAI 2022 · 85 citations) |
DFDC · AUC (%) (higher is better)
4 papers, 2021–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 73.5 | Lips Don't Lie: A Generalisable and Robust Approach To Face Forgery Detection (CVPR 2021 · 501 citations) |
| 2022 | 80.4 | Delving into Sequential Patches for Deepfake Detection (NeurIPS 2022 · 21 citations) |
DFEW · WAR (higher is better)
4 papers, 2023–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 69.25 | Rethinking the Learning Paradigm for Dynamic Facial Expression Recognition (CVPR 2023 · 107 citations) |
| 2025 | 77.45 | AVF-MAE++: Scaling Affective Video Facial Masked Autoencoders via Efficient Audio-Visual Self-Supervised Learning (CVPR 2025 · 7 citations) |
FF++ · AUC (higher is better)
4 papers, 2021–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 99.9 | Improving the Efficiency and Robustness of Deepfakes Detection Through Precise Geometric Features (CVPR 2021 · 172 citations) |
What has been tried
39 papers here report 445 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 4%, hurt 23%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add StyleGRU style-flow features to the model | CDF · AUC (%) | 85.7 -> 89.0 | Helped | Exploiting Style Latent Flows for Generalizing Deepfake Video Detection (CVPR 2024 · 77 citations) |
| add the spatial module | Cross-dataset average (CDF, DFDC, FSh, DFo, WDF) · video-level AUC (%) | 87.0(-5.3) -> 92.3 | Helped | Towards More General Video-based Deepfake Detection through Facial Component Guided Adaptation for Foundation Model (CVPR 2025 · 23 citations) |
| remove Global Encoder | LipSync detection · ACC | 95.27 -> 95.07 | Hurt | Lips Are Lying: Spotting the Temporal Inconsistency between Audio and Visual in Lip-Syncing DeepFakes (NeurIPS 2024 · 18 citations) |
| Replace vector-based GAT with MGNN on matrixial graph features | AVEC 2014 · RMSE | 8.01 -> 6.62 | Helped | DepMGNN: Matrixial Graph Neural Network for Video-based Automatic Depression Assessment (AAAI 2025 · 14 citations) |
| Remove feature normalization | FakeAVCeleb and AV-Deepfake1M · AUC | 94.6 -> 92.4; 85.9 -> 84.8 | Hurt | Circumventing Shortcuts in Audio-visual Deepfake Detection Datasets with Unsupervised Learning (CVPR 2025 · 9 citations) |
| Add diffusion reconstruction to Hybrid ViT | DVF average · AUC (%) | +1.7% AUC score | Helped | On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection (NeurIPS 2024 · 8 citations) |
| Scale the model from Base to Huge | MAFW (11-class) · WAR | 57.50 -> 60.24 | Helped | AVF-MAE++: Scaling Affective Video Facial Masked Autoencoders via Efficient Audio-Visual Self-Supervised Learning (CVPR 2025 · 7 citations) |
| Add the consecutive correlate module (CCM) to the real-face-pretrained ViT baseline | Celeb-DF, DFDC, and FFIW · frame-wise AUC (%) | 83.41 -> 86.30; 80.73 -> 82.99; 84.40 -> 87.36 | Helped | Face Forgery Video Detection via Temporal Forgery Cue Unraveling (CVPR 2025 · 6 citations) |
| add global temporal information modeling to local phonetic-level speech representations | FF++ cross-manipulation Avg · AUC (%) | 88.7 -> 97.6 | Helped | SpeechForensics: Audio-Visual Speech Representation Learning for Face Forgery Detection (NeurIPS 2024 · 6 citations) |
| Add ISSM and LADM | DFEW · UAR | 57.00 -> 61.19 | Helped | Lifting Scheme-Based Implicit Disentanglement of Emotion-Related Facial Dynamics in the Wild (AAAI 2025 · 2 citations) |
| remove the biometric leakage MLPs | NVIDIA VC, 3DFaceShop · AUC | 0.827 to 0.966 | Helped | Unmasking Puppeteers: Leveraging Biometric Leakage to Expose Impersonation in AI-Based Videoconferencing (NeurIPS 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 10 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- DFEW · 2 of 10 papers
- MAFW · 2 of 10 papers
- CREMA-D · 2 of 10 papers
- RAVDESS · 2 of 10 papers
- DFDC · 2 of 10 papers
- FaceForensics++ c23 · 2 of 10 papers
Baselines
- FTCN · 3 of 10 papers
- AltFreezing · 3 of 10 papers
- TALL · 3 of 10 papers
- Xception · 2 of 10 papers
- EfficientViT · 2 of 10 papers
- SBI · 2 of 10 papers
- NACO · 2 of 10 papers
- LAA-Net · 2 of 10 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 28 papers; from OpenAlex and the papers' affiliations).
- University of Science and Technology of China · 2 papers
- Zhejiang University · 2 papers
- NVIDIA · 2 papers
- Michigan State University · 2 papers
- Drexel University · 2 papers
Use it
Download BibTeX of 38 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Self-supervised video representation learning (167 papers)
- Video restoration with temporal priors (159 papers)
- Sign language and physiological video analysis (99 papers)
- Weakly supervised video anomaly detection (99 papers)
- Spatio-temporal predictive modeling (91 papers)
Elsewhere in the topic
- Layered video decomposition and dynamics (140 papers)
- Stochastic video prediction (125 papers)
- Physics-inspired motion prediction (113 papers)
- Matching-based video object segmentation (86 papers)
- Learned optical flow estimation (73 papers)
- Learned video compression (68 papers)
- Mask propagation for unsupervised video segmentation (67 papers)
- Video semantic and panoptic segmentation (67 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).