MultiID-2M: Large-Scale Paired Images for Multi-Person Identity Generation

MultiID-2M is a large-scale paired dataset constructed for multi-person scenarios in identity-consistent image generation. It provides diverse references for each identity to support the development of advanced diffusion-based models like WithAnyone. The dataset was created by jingchao-peng and was last updated on 2026-02-28.

Use Cases

• Training diffusion models for identity-consistent generation based on paired multi-person image references.
• Improving controllability over pose and expression in generated images based on the dataset's construction for multi-person scenarios.
• Mitigating 'copy-paste' artifacts in generated images based on the dataset's design purpose.

Strengths

• Designed for multi-person scenarios, a specific and complex task in image generation.
• Provides diverse references for each identity, a feature mentioned as enabling advanced model development.
• Associated with the WithAnyone research paper, indicating a research-oriented origin.

Limitations

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

Provenance

Source

jingchao-peng on Hugging Face

Collection Method

Specifically constructed for multi-person identity-consistent generation, likely from curated image sources.

Freshness

Last updated 2026-02-28 02:35:00; freshness should be verified.