HumanCentric Scenes AI: 296 AI-Generated Images Across Four Domains

296 AI-generated human-centric scenes form a multimodal benchmark for computer vision. The collection includes 85 surveillance-style images and 128 occupation-by-gender portraits. It was created by Nima0Kamali and last updated on Hugging Face in May 2026.

Use Cases

• Test object detection robustness based on low-resolution, noisy imagery mimicking CCTV footage.
• Evaluate algorithmic bias based on a balanced set of occupation and gender portraits.
• Benchmark multimodal AI systems based on the described human-centric scene categories.
• Train synthetic data classifiers based on AI-generated imagery from specific domains like surveillance or portraits.

Strengths

• Contains 296 total AI-generated images across multiple domains.
• Includes a specific subset of 85 images designed to test detection under heavy compression and low-light noise.
• Features a structured subset of 128 images with a paired design for occupation and gender analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the full dataset is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality and other subset details require manual inspection after download.

Provenance

Source

Hugging Face dataset by author Nima0Kamali.

Collection Method

Images were generated using AI tools, with specific mention of Midjourney for one subset.

Freshness

Last updated 2026-05-29 10:37:23; freshness should be verified.