Synthetic Graphic Designs with Multi-Conditional Annotations

400,000 synthetic graphic design samples, each annotated with multiple conditions for generation tasks. The dataset includes primary and secondary visual elements with spatial and semantic annotations, plus textual elements with layout information. It was created by HuiZhang0812 and was last updated in June 2025.

Use Cases

• Train models to generate layouts from textual slogans and layout information.
• Develop conditional image synthesis models using primary subject images and secondary decorative objects as inputs.
• Research spatial relationships in design by leveraging annotations for secondary visual elements.
• Build models for graphic design generation conditioned on a combination of visual and textual element annotations.

Strengths

• 400,000 high-quality synthetic samples.
• Each sample has rich multi-conditional annotations covering visual and textual elements.

Limitations

• All data is synthetic, which may not fully capture the complexity of real-world design processes.
• The dataset size of 400,000 samples may be insufficient for training very large generative models without augmentation.

Provenance

Source

Hugging Face dataset repository by HuiZhang0812.

Collection Method

Synthetically generated.

Freshness

Last updated June 2025.