Awesome Deepfakes Detection
A list of Deepfakes Detection datasets, tools, papers and code. If this list help you in your research, a star is my pleasure.
If you want to contribute to this list, welcome to send me a pull request or contact me :) .
This repo only collect papers related to Deepfake Detection. If you are also interested in Deepfakes, please refer to: Awesome Deepfakes.
Contents
Benchmark
For better comparison and research purpose, we also collect the benchmark of all the SOTA methods we can get on Celeb-DF, DFDC and FaceForensic++ datasets in video-level. We use AUC score (%) as the metrics. The results are presented as follows. Results in italics indicate they were conducted in Yuezun Li or Alexandros Haliassos, not in their original paper.
Celeb-DF(v2) | DFDC | FaceForensic++ | Year | Note | |
---|---|---|---|---|---|
Two-Stream | 53.8 | 61.4 | 70.7 | 2017 | FF++ only on DF subset. Use provided pre-trained model. |
Meso4 | 54.8 | 75.3 | 84.7 | 2018 | Same as above. |
MesoInception4 | 53.6 | 73.2 | 83.0 | 2018 | Same as above. |
FWA | 56.9 | 72.7 | 80.1 | 2018 | Same as above. |
DSP-FWA | 64.6 | 75.5 | 93.0 | 2018 | Same as above. |
VA-MLP | 55.0 | 61.9 | 66.4 | 2019 | Same as above. |
VA-LogReg | 55.1 | 66.2 | 78.0 | 2019 | Same as above. |
Xception-raw | 48.2 | 49.9 | 99.7 | 2019 | Same as above. |
Xception-c23 | 65.3 | 72.2 | 99.7 | 2019 | Same as above. |
Xception-c40 | 65.5 | 69.7 | 95.5 | 2019 | Same as above. |
Multi-Task | 54.3 | 53.6 | 76.3 | 2019 | Same as above. |
CapsuleNet | 57.5 | 53.3 | 96.6 | 2019 | Same as above. |
CNN-Spot | 75.6 | 72.1 | 65.7 | 2019 | FF++ only on FaceShifter HQ subset. All pretrained on FF++. |
Face X-ray | 79.5 | 65.5 | 92.8 | 2019 | Same as above. |
CNN-RNN | 69.8 | 68.9 | 80.8 | 2019 | Same as above. |
LipsForensics | 82.4 | 73.5 | 97.1 | 2020 | Same as above. |
Two-Branch | 76.7 | - | 93.2 | 2020 | - |
Patch-based | 69.6 | 65.6 | 57.8 | 2020 | Same as above. |
FD2Net | - | 66.09 | 99.45 | 2020 | - |
LRNet | 56.9 | - | 99.9 | 2021 | Trained on FF++. |
Datasets
Video Datasets
- UADFV: "Exposing Deep Fakes Using Inconsistent Head Poses." Paper
- EBV: "In Ictu Oculi: Exposing AI Generated Fake Face Videos by Detecting Eye Blinking." Paper Download
- Deepfake-TIMIT: "DeepFakes: a New Threat to Face Recognition? Assessment and Detection." Paper Download
- DFFD: "On the Detection of Digital Face Manipulation." Paper Download
- DeepfakeDetection: Download
- Celeb-DF (v1): "Celeb-DF: A Large-scale Challenging Dataset for DeepFake Forensics." Paper Download
- Celeb-DF (v2): "Celeb-DF: A Large-scale Challenging Dataset for DeepFake Forensics." Paper Download
- DFDC: "The DeepFake Detection Challenge (DFDC) Dataset." Paper Download
- FaceForensic++: "FaceForensics++: Learning to Detect Manipulated Facial Images." Paper Download
- FFIW-10K: "Face Forensics in the Wild." Paper Download
- Deeper Forensic-1.0: "DeeperForensics-1.0: A Large-Scale Dataset for Real-World Face Forgery Detection." Paper Download
- Wild Deepfake: "WildDeepfake: A Challenging Real-World Dataset for Deepfake Detection." Paper Download
- ForgeryNet: "ForgeryNet: A Versatile Benchmark for Comprehensive Forgery Analysis." Paper Download
Real Videos | Fake Videos | Year | Note | |
---|---|---|---|---|
UADFV | 49 | 49 | 2018 | focus on head pose |
EBV | - | 49 | 2018 | focus on eye blinking |
Deepfake-TIMIT | 320 | 640 | 2018 | GAN-Based methods |
DFFD | 1,000 | 3000 | 2019 | mutiple SOTA methods |
DeepfakeDetection | 363 | 3,608 | 2019 | collect from actors with publicly available generation methods |
Celeb-DF (v2) | 590 | 5639 | 2019 | high quality |
DFDC | 23,564 | 104,500 | 2019 | DFDC competition on Kaggle |
FaceForensic++ | 1,000 | 5,000 | 2019 | five different generation methods |
FFIW-10K | 10,000 | 10,000 | 2019 | mutiple faces in one frame |
DeeperForensics-1.0 | 50,000 | 10,000 | 2020 | add real-world perturbations |
Wild-Deepfake | 3,805 | 3,509 | 2021 | collect from Internet |
ForgeryNet | 99,630 | 121,617 | 2021 | 8 video-level generation methods, add perturbations |
Image Datasets
- DFFD: "On the Detection of Digital Face Manipulation." Paper Download
- FFHQ: "A Style-Based Generator Architecture for Generative Adversarial Networks." Paper Download
- iFakeFaceDB: "GANprintR: Improved Fakes and Evaluation of the State of the Art in Face Manipulation Detection." Paper Download
- 100k Faces Generated by AI (Online): Download
- DFGC: "DFGC 2021: A DeepFake Game Competition." Paper Dowload
- ForgeryNet: "ForgeryNet: A Versatile Benchmark for Comprehensive Forgery Analysis." Paper Download
Real Images | Fake Images | Year | Note | |
---|---|---|---|---|
DFFD | 58,703 | 240,336 | 2019 | mutiple SOTA methods |
FFHQ | - | 70,000 (GAN-based) | 2019 | generated by GAN-based methods |
iFakeFaceDB | - | 87,000 (StyleGAN) | 2020 | generated by StyleGAN |
100k Faces | - | 100,000 (StyleGAN) | 2021 | generated by StyleGAN |
DFGC | 1,000 | N*1,000 | 2021 | DFGC 2021 competition, fake images generated by users |
ForgeryNet | 1,438,201 | 1,457,861 | 2021 | 7 image-level generation methods, add perturbations |
Competition
Name | Link | Year | Note |
---|---|---|---|
Deepfake Detection Challenge | Website | 2019 | 1. video-level detection. 2. use DFDC datasets. 3. the first worldwide competition. 4. more than 2,000 teams participated. |
DeepForensics Challenge | Website | 2020 | 1. video-level detection. 2. use DeeperForensics-1.0 datasets. 3. simulates real-world scenarios. |
Deepfake Game Competition | Website | 2021 | 1. both image-level generation and video-level detection track. 2. use Celeb-DF(v2) datasets. |
Face Forgery Analysis Challenge | Website | 2021 | 1. both image-level and video-level detection track. 2. addtional temporal localization track. 3. use ForgeryNet dataset. |
Tools
Papers
CVPR
- "ForgeryNet: A Versatile Benchmark for Comprehensive Forgery Analysis", CVPR 2021: Paper Github
- "Representative Forgery Mining for Fake Face Detection", CVPR 2021: Paper Github
- "MagDR: Mask-Guided Detection and Reconstruction for Defending Deepfakes", CVPR 2021: Paper
- "Improving the Efficiency and Robustness of Deepfakes Detection Through Precise Geometric Features", CVPR 2021: Paper Github
- "Multi-Attentional Deepfake Detection", CVPR 2021: Paper Github
- "Lips Don't Lie: A Generalisable and Robust Approach To Face Forgery Detection", CVPR 2021: Paper
- "Spatial-Phase Shallow Learning: Rethinking Face Forgery Detection in Frequency Domain", CVPR 2021: Paper
- "Frequency-Aware Discriminative Feature Learning Supervised by Single-Center Loss for Face Forgery Detection", CVPR 2021: Paper
- "Generalizing Face Forgery Detection With High-Frequency Features", CVPR 2021: Paper
- "Face Forgery Detection by 3D Decomposition", CVPR 2021: Paper
- "Global Texture Enhancement for Fake Face Detection in the Wild", CVPR 2020: Paper
- "On the Detection of Digital Face Manipulation", CVPR 2020: Paper Github
- "Face X-Ray for More General Face Forgery Detection", CVPR 2020: Paper
- CNN-generated images are surprisingly easy to spot... for now", CVPR 2020: Paper Github
- "FReTAL: Generalizing Deepfake Detection using Knowledge Distillation and Representation Learning", CVPR Workshop 2021: Paper Github
- "OC-FakeDect: Classifying Deepfakes Using One-class Variational Autoencoder", CVPR Workshop 2020: Paper
- "Exposing DeepFake Videos By Detecting Face Warping Artifacts", CVPR Workshop 2019: Paper
- "Recurrent Convolutional Strategies for Face Manipulation Detection in Videos", CVPR Workshop 2019: Paper
- "Two-Stream Neural Networks for Tampered Face Detection", CVPR Workshop 2017: Paper
ICCV
- "KoDF: A Large-scale Korean DeepFake Detection Dataset", ICCV 2021: Paper
- "Artificial Fingerprinting for Generative Models: Rooting Deepfake Attribution in Training Data", ICCV 2021: Paper
- "Exploring Temporal Coherence for More General Video Face Forgery Detection", ICCV 2021: Paper
- "ID-Reveal: Identity-aware DeepFake Video Detection", ICCV 2021: Paper Github
- "Joint Audio-Visual Deepfake Detection", ICCV 2021: Paper
- "Learning Self-Consistency for Deepfake Detection", ICCV 2021: Paper
- "OpenForensics: Large-Scale Challenging Dataset For Multi-Face Forgery Detection And Segmentation In-The-Wild", ICCV 2021: Paper
- "Attributing Fake Images to GANs: Learning and Analyzing GAN Fingerprints", ICCV 2019: Paper Github
ECCV
- "Thinking in Frequency: Face Forgery Detection by Mining Frequency-aware Clues", ECCV 2020: Paper
- "Two-branch Recurrent Network for Isolating Deepfakes in Videos", ECCV 2020: Paper
ICML
IJCAI
- "Detecting Deepfake Videos with Temporal Dropout 3DCNN", IJCAI 2021: Paper
- "Dynamic Inconsistency-aware DeepFake Video Detection", IJCAI 2021: Paper
- "An Examination of Fairness of AI Models for Deepfake Detection", IJCAI 2021: Paper
- "Beyond the Spectrum: Detecting Deepfakes via Re-Synthesis", IJCAI 2021: Paper Github
- "FakeSpotter: A Simple yet Robust Baseline for Spotting AI-Synthesized Fake Faces", IJCAI 2020: Paper
AAAI
- "Dual Contrastive Learning for General Face Forgery Detection", AAAI 2022: Paper
- "CMUA-Watermark: A Cross-Model Universal Adversarial Watermark for Combating Deepfakes", AAAI 2022: Paper
- "ADD: Frequency Attention and Multi-View based Knowledge Distillation to Detect Low-Quality Compressed Deepfake Images", AAAI 2022: Paper
- "Domain General Face Forgery Detection by Learning to Weight", AAAI 2021: Paper Github
- "Local Relation Learning for Face Forgery Detection", AAAI 2021: Paper
NIPS
- "WaveFake: A Data Set to Facilitate Audio Deepfake Detection", NIPS 2021: Paper
- "AOT: Appearance Optimal Transport Based Identity Swapping for Forgery Detection", NIPS 2020: Paper
ICLR
- "Responsible Disclosure of Generative Models Using Scalable Fingerprinting", ICLR 2022: Paper
ACM MM
- "Evaluation of an Audio-Video Multimodal Deepfake Dataset using Unimodal and Multimodal Detectors", ACM MM 2021: Paper
- "Spatiotemporal Inconsistency Learning for DeepFake Video Detection", ACM MM 2021: Paper
- "Video Transformer for Deepfake Detection with Incremental Learning", ACM MM 2021: Paper
- "Metric Learning for Anti-Compression Facial Forgery Detection", ACM MM 2021: Paper
- "CoReD: Generalizing Fake Media Detection with Continual Representation using Distillation", ACM MM 2021: Paper
- "FakeTagger: Robust Safeguards against DeepFake Dissemination via Provenance Tracking", ACM MM 2021: Paper
- "FakeAVCeleb: A Novel Audio-Video Multimodal Deepfake Dataset", ACM MM 2021 Workshop: Paper Github
- "Not made for each other- Audio-Visual Dissonance-based Deepfake Detection and Localization", ACM MM 2020: Paper Github
- "Sharp Multiple Instance Learning for DeepFake Video Detection", ACM MM 2020: Paper
- "DeepRhythm: Exposing DeepFakes with Attentional Visual Heartbeat Rhythms", ACM MM 2020: Paper
- "Emotions Don't Lie: An Audio-Visual Deepfake Detection Method using Affective Cues", ACM MM 2020: Paper
ICME
- "DEEPFAKE VIDEOS DETECTION USING SELF-SUPERVISED DECOUPLING NETWORK", ICME 2021: Paper
- "DLFMNET: END-TO-END DETECTION AND LOCALIZATION OF FACE MANIPULATION USING MULTI-DOMAIN FEATURES", ICME 2021: Paper Github
- "DEFAKEHOP: A LIGHT-WEIGHT HIGH-PERFORMANCE DEEPFAKE DETECTOR", ICME 2021: Paper Github
- "FSSPOTTER: Spotting Face-Swapped Video by Spatial and Temporal Clues", ICME 2020: Paper
TPAMI
- "DeepFake Detection Based on Discrepancies Between Faces and their Context", TPAMI 2021: Paper
- "FakeCatcher: Detection of Synthetic Portrait Videos using Biological Signals", TPAMI 2021: Paper
TIFS
- "Detection of Fake and Fraudulent Faces via Neural Memory Networks", TIFS 2021: Paper
- "Preventing DeepFake Attacks on Speaker Authentication by Dynamic Lip Movement Analysis", TIFS 2021: Paper
Other
- "FakeCatcher: Detection of Synthetic Portrait Videos using Biological Signals", TPAMI 2020: Paper
- "One Detector to Rule Them All: Towards a General Deepfake Attack Detection Framework", WWW 2021: Paper
- "How Do the Hearts of Deep Fakes Beat? Deep Fake Source Detection via Interpreting Residuals with Biological Signals", IJCB 2020: Paper
- "Deepfake Detection using Spatiotemporal Convolutional Networks", arxiv: Paper Github
- "A Convolutional LSTM based Residual Network for Deepfake Video Detection", arxiv: Paper
- "Spatio-temporal Features for Generalized Detection of Deepfake Videos", submitted to CVIU: Paper
- "Exploiting Visual Artifacts to Expose Deepfakes and Face Manipulations", WACVW 2019: Paper
- "Interpretable and Trustworthy Deepfake Detection via Dynamic Prototypes", WACV 2021: Paper
- "MesoNet: a Compact Facial Video Forgery Detection Network", WIFS 2018: Paper Github
- "Multi-task Learning For Detecting and Segmenting Manipulated Facial Images and Videos", BATS 2019: Paper
- "Use of a Capsule Network to Detect Fake Images and Videos", arxiv: Paper
- "What makes fake images detectable? Understanding properties that generalize", arxiv: Paper
- "PRRNet: Pixel-Region relation network for face forgery detection", Pattern Recognition 2021: Paper
- "ForensicTransfer: Weakly-supervised Domain Adaptation for Forgery Detection", arxiv: Paper
- "Identity-Driven DeepFake Detection", CoRR 2020: Paper
- "Exposing Fake Faces Through Deep Neural Networks Combining Content and Trace Feature Extractors", IEEE Access 2021: Paper
- "FaceGuard: Proactive Deepfake Detection", CoRR 2021: Paper
- "DeepFake-o-meter: An Open Platform for DeepFake Detection", SP Workshops 2021: Paper
- "DeepfakeUCL: Deepfake Detection via Unsupervised Contrastive Learning", IJCNN 2021: Paper
- "M2TR: Multi-modal Multi-scale Transformers for Deepfake Detection", CoRR 2021: Paper