Vision Sample Data

A curated dataset of labeled images for testing vision classification features in LlamaFarm Designer.

Folder Structure

├── cats/              # 12 images of domestic cats
├── firetrucks/        # 12 images of fire trucks/engines
├── sunflowers/        # 12 images of sunflowers
├── pizza/             # 12 images of pizza
└── none-of-the-above/ # 15 miscellaneous images (landmarks, nature, instruments, etc.)

What This Is For

This dataset is designed for testing LlamaFarm Designer's vision classification flow. The images are organized into clear categories so you can verify that vision models correctly identify and classify image content.

The none-of-the-above/ folder provides negative examples — images that don't belong to any of the four main categories — useful for testing classification confidence and rejection behavior.

How to Use

1. Clone or download this repo
2. In LlamaFarm Designer, point your vision classification pipeline at images from this repo
3. Use the folder names as ground-truth labels to evaluate classification accuracy

You can also reference images directly via GitHub raw URLs:

https://raw.githubusercontent.com/llama-farm/vision-sample-data/main/cats/cats_01.jpg

Image Details

• All images resized to max 800px wide
• JPEG format
• Sourced from Wikimedia Commons (free/CC/public domain licensed)
• See SOURCES.md for per-image attribution

Note

This is a standalone sample data repo — it is not bundled as a LlamaFarm addon. Users link to it directly for testing purposes.

License

All images are sourced from Wikimedia Commons under various free licenses (CC BY, CC BY-SA, CC0/Public Domain). See SOURCES.md for individual image sources and their respective licenses.