Open-Personix: Person-Centered Multimodal Captions and Metadata

Open-Personix is a person-centered multimodal dataset of fewer than 1,000 records maintained by Poralus and updated in March 2026. It provides structured JSON entries containing relative image paths, natural-language captions, and descriptive person-specific annotations.

Use Cases

• Training caption generation models using the natural-language caption field
• Performing text-based filtering of person samples using descriptive annotation fields
• Benchmarking multimodal retrieval systems using relative image paths and metadata

Strengths

• MIT license allows for commercial and research use
• Includes both unstructured captions and structured descriptive annotations
• Native support for Polars and Pandas libraries

Limitations

• Small sample size of under 1,000 records
• Requires external image assets as only relative paths are provided
• Specific schema for descriptive annotation fields is not explicitly documented

Provenance

Source

Poralus

Freshness

Last updated March 2026.

Geography

United States