Balanced Dataset for AI-Generated Versus Real Image Classification

A balanced collection of AI-generated and real human-captured images for classification tasks. The dataset was created by Parveshiiii and was last updated on September 25, 2025. It contains two classes: '0' for AI-generated images and '1' for real images.

Use Cases

• Train binary classifiers to predict the 'class' label (0 or 1) from image pixel data.
• Evaluate model robustness on the balanced 'AI-generated' and 'Real' image classes.
• Benchmark feature extraction methods for distinguishing synthetic from authentic images.
• Fine-tune vision transformers using the labeled 'class' attribute for each image.

Strengths

• Dataset is explicitly balanced across the two primary classes.
• Contains two clearly defined classes: '0' for AI-generated and '1' for real images.

Limitations

• Specific row count, image resolution, and source details are unknown.
• Potential for label noise or misclassification between synthetic and real categories.
• Geographic and subject diversity of the 'real' images is unspecified.

Provenance

Source

Parveshiiii on Hugging Face.

Freshness

Last updated on 2025-09-25.