brand-genome/brand-guidelines-pdfs

BrandGuide / Brand Genome Dataset

This dataset contains extracted brand guidelines for brand-level feature engineering and multimodal brand analysis.

Overview

BrandGuide is a large-scale collection of brand guidelines paired with corresponding brand assets. According to the paper, it covers 2,683 brands, spans 80 sectors, 103 regions, and 28 languages, with roughly 1M images and text assets collected over 2014–2025. The dataset is designed for interpretable ML and expert-aligned feature discovery in high-stakes brand compliance settings.

Dataset Contents

This repository is organized around brand-specific guideline PDFs:

• pdfs/ — brand guideline PDF files
• brand_to_pdf_manifest.json — mapping from brand name to the PDF file name(s) associated with that brand

A single brand may have multiple PDFs.

Intended Use

The dataset is meant for research on:

• brand guideline understanding
• interpretable feature engineering
• brand compliance and consistency checking
• multimodal brand analysis
• expert-aligned AI systems

Example Entry Structure

Each brand entry may include:

• brand metadata
• one or more guideline PDFs
• extracted guideline text
• associated brand imagery or other assets, when available

Data Statistics

• Brands: 2,683
• Sectors: 80
• Regions: 103
• Languages: 28
• Temporal coverage: 2014–2025
• Visual assets: ~1M images

File Layout

dataset_root/
├── pdfs/
│   ├── brand_a_guidelines.pdf
│   ├── brand_b_styleguide_part1.pdf
│   ├── brand_b_styleguide_part2.pdf
│   └── ...
├── brand_to_pdf_manifest.json
└── README.md

Notes

• The dataset is intended for non-commercial research use.
• Intellectual property rights remain with the original brand owners and guideline authors.
• Users should verify provenance and comply with the original source terms for each document.

Citation

If you use this dataset in research, please cite the associated paper on FEST (Feature Engineering with Self-evolving Trees) and BrandGuide.

@misc{khurana2026bridgingexpertknowledgeautomated,
      title={Bridging Expert Knowledge and Automated Feature Engineering via Self-Evolution}, 
      author={Varun Khurana and Vijval Ekbote and Vashu Chauhan and Yaman Kumar Singla and Rajiv Ratn Shah and Balaji Krishnamurthy},
      year={2026},
      eprint={2606.08800},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2606.08800}, 
}