gauravs101/synthetic-greeting-cards

Synthetic Greeting Cards Dataset (Synth-GCD)

A modern, fully open synthetic alternative to the proprietary Greeting Cards Dataset (GCD) described in the paper "Weakly Supervised Annotations for Multi-modal Greeting Cards Dataset".

This dataset contains high-quality AI-generated greeting card illustrations with corresponding short messages, designed for research in multimodal classification, retrieval, and generation.

Dataset Summary

Motivation

The original GCD dataset (3,700+ images) is proprietary and not publicly available. This synthetic version was created so researchers can freely experiment with multimodal greeting-card understanding without legal or access restrictions.

Dataset Structure

synth_gcd/
├── synth_gcd_50/          # First batch (numbered + randomized names)
│   ├── 01_christmas.jpg
│   ├── 02_valentines.jpg
│   ├── ...
│   ├── lm1.jpg ... lm10.jpg
│   └── ...
├── synth_gcd_extra/       # Second batch (grouped naming)
│   ├── sm1.jpg ... sm10.jpg
│   ├── em1.jpg ... em10.jpg
│   ├── pm1.jpg ... pm10.jpg
│   ├── km1.jpg ... km10.jpg
│   └── extra1.jpg, extra2.jpg
└── README.md

Theme Coverage

The images cover a wide range of greeting card themes, including (but not limited to):

Holidays & Seasons

• —Christmas, New Year, Easter, Halloween, Diwali, Thanksgiving, Winter Greetings

Special Occasions

• —Birthday, Anniversary, Wedding, Graduation, New Baby, Retirement, Housewarming, Engagement

Messages & Sentiments

• —Thank You, Thinking of You, Friendship, Encouragement, Get Well, Sorry, Good Luck, Farewell
• —New Beginnings, Hope, Peace, Strength, Kindness, Growth, Journey

How the Data Was Generated

• —All images were generated using modern open text-to-image models.
• —Prompts were carefully crafted to produce greeting-card style illustrations (flat, watercolor, soft cartoon).
• —Each image includes a short, natural cover message appropriate for the theme.
• —No real commercial greeting cards were scraped or copied.

Intended Uses

• —Multimodal classification (image + text)
• —Image-to-text / text-to-image research
• —Style transfer and illustration generation
• —Dataset construction and benchmarking for greeting-card domains
• —Educational and research purposes

Limitations

• —Synthetic data – visual style is consistent with modern AI illustration rather than traditional printed cards.
• —Text is rendered inside the image (not provided as separate text files).
• —Relatively small scale compared to the original proprietary GCD (~100 images vs 3,700+).

License

Apache License 2.0

You are free to use, modify, and distribute this dataset for research and commercial purposes, provided you include the original license and attribution.

Citation

If you use this dataset, please consider citing the original GCD paper for context:

@article{gcd2022,
  title={Weakly Supervised Annotations for Multi-modal Greeting Cards Dataset},
  year={2022}
}

And optionally mention this synthetic version as:

Synthetic Greeting Cards Dataset (Synth-GCD) – open alternative generated for research reproducibility.

Contact / Contributions

Feel free to open issues or pull requests if you expand the dataset with more categories, higher resolution, or separate text annotations.