DeeperForensics-1.0: A Large-Scale Dataset for Real-World Face Forgery Detection

By Liming Jiang, Wayne Wu, Ren-Cang Li, Chen Qian, Chen Change Loy. Computer Vision and Pattern Recognition, 2020

617 citations recorded by the source corpus.

We present our on-going effort of constructing a large- scale benchmark for face forgery detection. The first version of this benchmark, DeeperForensics-1.0, represents the largest face forgery detection dataset by far, with 60, 000 videos constituted by a total of 17.6 million frames, 10 times larger than existing datasets of the same kind. Extensive real-world perturbations are applied to obtain a more challenging benchmark of larger scale and higher diversity. All source videos in DeeperForensics-1.0 are carefully collected, and fake videos are generated by a newly proposed end-to-end face swapping framework. The quality of generated videos outperforms those in existing datasets, validated by user studies. The benchmark features a hidden test set, which contains manipulated videos achieving high deceptive scores in human evaluations. We further contribute a comprehensive study that evaluates five representative detection baselines and make a thorough analysis of different settings.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/1456b01982bed948d578a69f4c67d14a110e68fd

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Learning Self-Consistency for Deepfake Detection (2020, methodology): “0 (DFR) [15] modifies the pristine videos in FF++ with new face IDs and more advanced techniques.”
Self-supervised Learning of Adversarial Example: Towards Good Generalizations for Deepfake Detection (2022, methodology): “0) [20], which consists of over 11,000 fake videos generated by their DFVAE [20] method.”
Interpretable and Trustworthy Deepfake Detection via Dynamic Prototypes (2021, neutral): “5). distortions [27] (Figure 5).”
End-to-End Reconstruction-Classification Learning for Face Forgery Detection (2022, background): “Considering the ubiquity of image processing on social media, we investigate the performance under several perturbations suggested by [15, 17], i.”
FakeAVCeleb: A Novel Audio-Video Multimodal Deepfake Dataset (2021, methodology): “Keeping in mind the essence of datasets, researchers have also proposed different deepfake datasets, generated using the latest deepfake generation methods [9, 11, 15, 19, 22, 32] to help researchers train and evaluate their deepfake detection methods.”
Generalizing Face Forgery Detection with High-frequency Features (2021, background): “As face manipulation techniques [4, 24, 59] spring up along with the breakthrough of deep generators [27, 20], face forgery detection becomes an arousing research topic.”
Lips Don't Lie: A Generalisable and Robust Approach to Face Forgery Detection (2020, methodology): “Other datasets in our experiments include DeeperForensics [30] and FaceShifter [34], each featuring an improved face swapping algorithm applied to the real videos from FF++.”
One Shot Face Swapping on Megapixels (2021, background): “Besides, it is time consuming and difficult for subject-specific methods to train specific models for distinct pairs of faces [14, 46, 45, 29, 20, 55, 26].”
A Continual Deepfake Detection Benchmark: Dataset, Methods, and Essentials (2022, background): “To address this issue, many deepfake detection datasets (e.g.,[40, 44, 46, 21, 64, 19, 31]) and techniques (e.g., [87, 4, 80, 57, 56, 5]) are proposed.”
The DeepFake Detection Challenge Dataset (2020, background): “Our hypothesis is that GAN-like methods work well in limited settings with even lighting, such as news rooms, interviews, or controlled-capture videos as in [14], but do not work well automatically (yet).”
Deepfakes generation and detection: state-of-the-art, open challenges, countermeasures, and way forward (2021, background): “DeeperForensics (DF): Another Large-Scale dataset for deepfake detection containing 50,000 original and 10,000 manipulated videos is built-in [297].”
Emotions Don't Lie: An Audio-Visual Deepfake Detection Method using Affective Cues (2020, neutral): “0 [31] Jan 2020 50,000 10,000 60,000 Actors DF ✓ –”
The DeepFake Detection Challenge (DFDC) Dataset. (2020, neutral): “Our hypothesis is that GAN-like methods work well in limited settings with even lighting, such as news rooms, interviews, or controlled-capture videos as in [14], but do not work well automatically (yet).”
Detecting Deepfakes with Self-Blended Images (2022, methodology): “0 [29] provide sophisticated deepfake videos, we do not adopt them in our cross-dataset evaluation because they generate deepfakes from the real videos of FF++ that is the same domain as used in training.”