The ForgeryNet dataset is a large and comprehensive benchmark built specifically for deepfake analysis. It contains 2.9 million images and 221,247 videos, covering 7 image-level and 8 video-level forgery manipulation methods from around the world. This dataset provides researchers with a rich resource to support four tasks at the image and video levels: image forgery classification, spatial forgery localization, video forgery classification, and temporal forgery localization. These tasks include image forgery identification from binary to multi-classification, as well as spatial and temporal localization of forged areas.

The scale and diversity of the ForgeryNet dataset make it the largest deep face forgery dataset currently available. Features are as follows:

1. Data-scale: 2.9 million images, 221,247 videos
2. Manipulations: 7 image-level methods, 8 video-level methods
3. perturbations: 36 individual perturbations and more mixed perturbations
4. annotations: 6.3 million classification labels, 2.9 million action region annotations, and 221,247 temporal forgery segment labels

Can be used for the following four tasks:

• Time forgery location: locate which time periods in which videos have been forged.

• Image forgery classification, including two-way (real/forged), three-way (real/forged vs. identity-replacing forgery method/forged vs. identity-preserving forgery method), and n-way (real vs. 15 forgery methods) classification.

• Spatial forgery localization,The manipulated regions of forged images are segmented compared with the corresponding real images.

• Video Forgery Classification,Redefine video-level forgery classification and process frames at random positions.

The dataset was jointly launched in 2021 by researchers from SenseTime Research, Beijing University of Posts and Telecommunications, Shanghai Artificial Intelligence Laboratory, School of Software, Beihang University, University of Science and Technology of China, and S-Lab, Nanyang Technological University.ForgeryNet: A Versatile Benchmark for Comprehensive Forgery Analysis".