Real vs Photoshop vs GAN vs Diffusion Faces: A Multi-Class Forgery Dataset

First multi-class face forgery dataset, containing images labeled as real, photoshopped, GAN-generated, and diffusion-generated. The dataset is hosted on Kaggle, but specific details on size, creation date, and authorship are not provided. Its primary purpose is to support research in detecting manipulated facial imagery from multiple modern generation techniques.

Use Cases

• Train classifiers to distinguish real faces from manipulated ones based on the four-class structure.
• Benchmark detection algorithms against multiple forgery types (Photoshop, GAN, Diffusion).
• Analyze visual artifacts specific to different image generation and editing methods.

Strengths

• It is described as the first multi-class face forgery dataset, covering four distinct categories.
• The dataset explicitly includes faces generated by modern AI techniques like GANs and diffusion models.

Limitations

• Row count and total dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Kaggle