metadata
language:
- en
license: apache-2.0
task_categories:
- text-generation
- image-to-text
- feature-extraction
pretty_name: WebUI
size_categories:
- 10K<n<100K
🌐 WebUI
A high-quality dataset for training AI models to understand and generate modern websites.
The dataset contains structured webpage samples including HTML, screenshots, UI element annotations, semantic labels, color palettes, fonts, layout metadata, and accessibility information. Each sample represents a complete webpage that can be used for web generation, UI understanding, or multimodal training. :contentReference[oaicite:0]{index=0}
✨ Features
- 📄 Complete HTML pages
- 🖼️ Full-page screenshots
- 📱 Desktop, tablet, and mobile layouts
- 🎨 Extracted color palettes
- 🔤 Font metadata
- 📦 Structured UI elements
- 📍 Bounding boxes
- ♿ Accessibility labels
- 🏷️ Page categories
- 🔗 Original source URLs
Example Uses
- Fine-tune LLMs for HTML generation
- Website recreation from screenshots
- UI parsing
- Layout understanding
- Frontend code generation
- Vision-language model training
- Accessibility research
Dataset Structure
Each record includes fields similar to:
{
"id": "...",
"url": "...",
"title": "...",
"page_type": "...",
"device_type": "...",
"html": "...",
"elements": [...],
"images": [...],
"colors": [...],
"fonts": [...],
"screenshot": "..."
}
Loading
from datasets import load_dataset
dataset = load_dataset("xlelords/webui")
print(dataset["train"][0])
Recommended Models
- Qwen 3 4B
- Gemma 3 4B
- Llama 3.2 3B
- SmolLM2
- Phi-4 Mini
- TinyLlama
Example Prompts
- Create a responsive SaaS landing page.
- Build a modern portfolio website.
- Generate a beautiful login screen.
- Design an AI dashboard.
- Create an ecommerce homepage.
- Generate a banking dashboard.
- Build a travel booking website.
- Make a glassmorphism pricing page.
Citation
If you use this dataset in your work, please cite the Hugging Face dataset repository.
License
Apache-2.0
Made for developers, researchers, and anyone building the next generation of AI-powered web design tools.