CIRCL/circl-ail-dataset-01-updated

---
license: cc-by-4.0
---
# circl-ail-dataset-01-updated

This repository contains an updated and enriched version of the [original CIRCL Images AIL Dataset](https://www.circl.lu/opendata/circl-ail-dataset-01/).

The updated dataset adds two kinds of derived metadata for each image:

- **LLM-generated image descriptions**
- **Open Alleged PhotoDNA fingerprints**

These records are intended to support research, analysis, and experimentation on image classification, content understanding, and similarity matching.

## Dataset contents

The repository includes processed JSON records derived from the original dataset.

### Open Alleged PhotoDNA fingerprints

The following files contain the [Open Alleged PhotoDNA](https://github.com/adulau/open-alleged-photodna) fingerprints:

- `folder_0.json`
- `folder_1_SECONDPASS.json`

These files store fingerprint data associated with the images from the original CIRCL Images AIL Dataset.

### LLM image descriptions

Each image is associated with a JSON record containing a machine-generated description produced by a large language model (LLM).

Each record includes:

- the image filename
- the model used
- the prompt sent to the model
- the generated response

Example:

```json
{
"image": "enormous-ugly-used-mixture.png",
"model": "qwen3-vl:32b",
"prompt": "Describe the image",
"response": "This image is a screenshot of the Spanish-language website ..."
}