Research map / Detection and segmentation / CNN features
Forgery detection
Detect manipulated faces and images with frequency, noise, and local anomaly cues.
46 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in CNN features, keyword cluster “local · context · semantic segmentation”
State History Benchmarks Tried Preprints Who Export
Where it stands
Few recent reports None of the tracked benchmarks has a report from 2025–26.
13 papers in 2023–24 and 14 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 4.0% to 4.0%.
No longer reported by papers here: Celeb-DF (AUC, last in 2023); FaceForensics++ HQ (AUC, last in 2022); FaceForensics++ LQ (ACC, last in 2022); FaceForensics++ LQ (AUC, last in 2022).
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
- 2017 · Action Unit Detection With Region Adaptation, Multi-Labeling Learning and Optimal Temporal Fusing (CVPR 2017 · 165 citations)
- 2018 · Learning Rich Features for Image Manipulation Detection (CVPR 2018 · 766 citations)
- 2019 · ManTra-Net: Manipulation Tracing Network for Detection and Localization of Image Forgeries With Anomalous Features (CVPR 2019 · 649 citations)
- 2020 · On the Detection of Digital Face Manipulation (CVPR 2020 · 722 citations)
- 2021 · Multi-Attentional Deepfake Detection (CVPR 2021 · 942 citations)
- 2022 · End-to-End Reconstruction-Classification Learning for Face Forgery Detection (CVPR 2022 · 394 citations)
- 2023 · TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization (CVPR 2023 · 237 citations)
- 2024 · Rethinking the Up-Sampling Operations in CNN-based Generative Network for Generalizable Deepfake Detection (CVPR 2024 · 200 citations)
- 2025 · MUN: Image Forgery Localization Based on M³ Encoder and UN Decoder (AAAI 2025 · 11 citations)
- 2026 · Fine-Grained DINO Tuning with Dual Supervision for Face Forgery Detection (AAAI 2026 · 4 citations)
Earliest papers here
- Action Unit Detection With Region Adaptation, Multi-Labeling Learning and Optimal Temporal Fusing (CVPR 2017 · 165 citations)
- Image Splicing Detection via Camera Response Function Analysis (CVPR 2017 · 51 citations)
- Learning Rich Features for Image Manipulation Detection (CVPR 2018 · 766 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, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 71.2 | On the Detection of Digital Face Manipulation (CVPR 2020 · 722 citations) |
| 2022 | 99.94 | End-to-End Reconstruction-Classification Learning for Face Forgery Detection (CVPR 2022 · 394 citations) |
| 2023 | 99.96 | Dynamic Graph Learning With Content-Guided Spatial-Frequency Relation Reasoning for Deepfake Detection (CVPR 2023 · 150 citations) |
FaceForensics++ HQ · AUC (higher is better)
4 papers, 2021–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 99.46 | Local Relation Learning for Face Forgery Detection (AAAI 2021 · 314 citations) |
FaceForensics++ LQ · ACC (higher is better)
4 papers, 2021–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 91.47 | Local Relation Learning for Face Forgery Detection (AAAI 2021 · 314 citations) |
FaceForensics++ LQ · AUC (higher is better)
4 papers, 2021–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 95.21 | Local Relation Learning for Face Forgery Detection (AAAI 2021 · 314 citations) |
What has been tried
36 papers here report 402 measured changes (a component added, removed or swapped, compared with a baseline): helped 80%, no clear effect 3%, hurt 14%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use w2 rather than w1 as the 2×2 NPR reference pixel | ForenSynths test set · Acc. | 92.5 -> 87.5 | Hurt | Rethinking the Up-Sampling Operations in CNN-based Generative Network for Generalizable Deepfake Detection (CVPR 2024 · 200 citations) |
| add Local Information Loss and Global Information Loss | FF++ (C23) · AUC | 0.966 to 0.983 | Helped | Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · 84 citations) |
| Add the heatmap branch | CDF2 · AUC (%) | 79.87 -> 95.40 | Helped | LAA-Net: Localized Artifact Attention Network for Quality-Agnostic and Generalizable Deepfake Detection (CVPR 2024 · 75 citations) |
| Add noise representation learning (NRL) | CASIA and NIST16 · AUC | CASIA: 82.6 -> 91.3; NIST16: 83.2 -> 99.8 | Helped | Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · 28 citations) |
| add the noise view to the RGB view | CASIA v1 · pixel-level AUC | 0.778 -> 0.865 | Helped | UnionFormer: Unified-Learning Transformer with Multi-View Representation for Image Manipulation Detection and Localization (CVPR 2024 · 26 citations) |
| Add a noise branch and SCAM | Columbia · AUC (%) | 96.0 -> 97.1 | Helped | MGQFormer: Mask-Guided Query-Based Transformer for Image Manipulation Localization (AAAI 2024 · 20 citations) |
| add aggressive feature aligned learning (AFAL) | GAP, including NAPs and Non-NAPs · AP@0.5 | 76.84 -> 91.89 | Helped | NAPGuard: Towards Detecting Naturalistic Adversarial Patches (CVPR 2024 · 20 citations) |
| Use IoUDCE loss instead of cross-entropy loss | Synthesized Dataset test · F1 | 0.9519 -> 0.9523 | Helped | MUN: Image Forgery Localization Based on M³ Encoder and UN Decoder (AAAI 2025 · 11 citations) |
| Add CatmullRom splines instead of traditional regression | CASIA · AUC | 68.5 -> 90.4 | Helped | CatmullRom Splines-Based Regression for Image Forgery Localization (AAAI 2024 · 11 citations) |
| Add feature decoupling instead of decoupling images | CDF (FF++ c40 cross-dataset) · AUC | 70.80 -> 72.45 | Helped | Critical Forgetting-Based Multi-Scale Disentanglement for Deepfake Detection (AAAI 2025 · 3 citations) |
| add FMN while keeping the backbone frozen | MVTec AD 2-shot · Instance AUROC | 85.43 -> 88.39 | Helped | DFM: Differentiable Feature Matching for Anomaly Detection (CVPR 2025 · 2 citations) |
| add vulnerability-type knowledge K1 | Tm119 · F1 | 11.9% | Helped | Suitable is the Best: Task-Oriented Knowledge Fusion in Vulnerability Detection (NeurIPS 2024 · 2 citations) |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 27 papers; from OpenAlex and the papers' affiliations).
- Chinese Academy of Sciences · 5 papers
- University of Chinese Academy of Sciences · 3 papers
- University of Science and Technology of China · 2 papers
- Zhejiang University · 2 papers
- Zhongguancun Laboratory · 2 papers
- Beihang University · 2 papers
Use it
Download BibTeX of 37 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
- Context modeling for semantic segmentation (241 papers)
- Local features and context in CNNs (170 papers)
- Context for hard detection conditions (133 papers)
- Domain-specific segmentation and detection (118 papers)
- Lane detection (32 papers)
Elsewhere in the topic
- Region-based and query-based detectors (183 papers)
- Convolution-transformer backbones (101 papers)
- Box regression and loss design (94 papers)
- Efficient DETR query design (85 papers)
- Boundary-aware instance segmentation (83 papers)
- Plain and high-resolution ViTs (69 papers)
- Arbitrary-shape scene text detection (61 papers)
- RGB salient object detection (61 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).