OverlayDataset: 499,249 Images with Object Layouts and Scene Captions

OverlayDataset is a large-scale vision-language collection of 499,249 images paired with dense object-level annotations, local prompts for objects, and global scene captions. It was created by dsrivastavv and last updated on Hugging Face in April 2026. The dataset is designed for training controllable image generation systems.

Use Cases

• Train image generation models based on object layouts and local prompts.
• Fine-tune vision-language models using dense object-level annotations.
• Develop scene understanding systems based on long and short global captions.
• Research controllable image synthesis based on the provided multi-level conditioning data.

Strengths

• Contains 499,249 images, providing a large-scale resource.
• Includes multiple annotation types per image: object-level annotations, local prompts, and global captions.
• Provides original source URLs and content hashes for traceability when available.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update date is in the future (2026-04-09).

Provenance

Source

Hugging Face user dsrivastavv.

Collection Method

Likely aggregated from various online sources, as indicated by the inclusion of original URLs.

Freshness

Last updated 2026-04-09 02:33:43