Real Versus AI Image Classification Corpus

US region binary classification dataset for training AI-image detectors, built by Zitacron from 17 public HuggingFace sources. The corpus contains real and AI-generated images, all with commercial licenses.

Use Cases

• Train binary classification models to detect AI-generated images using the dataset's image and label fields.
• Fine-tune vision models on the corpus for the specific AI Detection task category.
• Analyze feature distributions between Real Images and AI Generated categories to understand generation artifacts.

Strengths

• Built from 17 distinct public sources, providing diverse image origins.
• All constituent sources use commercially permissive licenses (CC BY 4.0, Apache 2.0, or MIT).
• Dataset is structured for the specific Task Categories of image Classification and binary Classification.

Limitations

• Unknown row count and column schema limits precise assessment of dataset scale and structure.
• Potential geographic bias indicated by the Regionus tag.
• Sample data and file formats are unavailable, hindering preliminary inspection.

Provenance

Source

HuggingFace, aggregated from 17 public sources.

Collection Method

Streaming-merged from constituent sources with no intermediate local storage.

Time Range

null

Freshness

Last updated 2026-03-21.

Geography

US (from tags)