Deep Fake Detection DFD Entire Original Dataset
Video dataset designed for the detection of deepfakes, containing a large collection of original and manipulated video sequences, useful for training and testing algorithms for detecting facial manipulations.
Approximately 3400 video files (original and manipulated), MP4 formats
MIT
Description
The dataset Deep Fake Detection DFD Entire Original Dataset includes approximately 3400 video files, including original and manipulated sequences, intended for training and evaluating deepfake detection models. The videos come from the official FaceForensics server, guaranteeing high data quality.
What is this dataset for?
- Training deep learning models to detect deepfake videos
- Comparative evaluation of algorithms for detecting facial manipulations
- Research in digital security and media integrity
Can it be enriched or improved?
The dataset can be enriched by adding new deepfake videos from other techniques, or by detailed annotations on the types of manipulations.
🔎 In summary
🧠 Recommended for
- Video vision researchers
- Security developers
- AI R&D teams
🔧 Compatible tools
- OpenCV
- FFmpeg
- TensorFlow
- PyTorch
- DeepFaceLab
💡 Tip
Use frame extraction and video augmentation techniques to improve model performance.
Frequently Asked Questions
Does this dataset contain manipulated and original videos?
Yes, it includes both original videos and their manipulated versions for the detection of deepfakes.
Does the license allow commercial use?
Yes, the MIT license allows free commercial use.
What is the format of video files?
The files are mostly in MP4 format, which is standard for video.
