---
license: cc-by-nc-nd-4.0
task_categories:
- visual-question-answering
size_categories:
- 1K<n<10K
---
# A Visual RAG Pipeline for Few-Shot Fine-Grained Product Classification

## Paper

Accepted at *The 12th Workshop on Fine-Grained Visual Categorization* ([FGVC12](https://sites.google.com/view/fgvc12)) at *IEEE/CVF Conference on Computer Vision and Pattern Recognition* ([CVPR](https://cvpr.thecvf.com/)) 2025.

## Overview

The **Retail Visual RAG Pipeline Dataset** is a subset of the *Retail-786k* ([Retail-786k](https://www.retail-786k.org/)) image dataset, supplemented with additional textual data per image.

### Data

- **Image Data:**

  - The images are cropped from scanned advertisement leaflets.
  - The image data is divided into `train` and `test` splits.

- **Product and Promotion Data:**

  Product Data:
  - Product data contains the targets: brand, product category, GTINs, product weight, and different sorts.
  - If a promotion covers a variety of different types/flavors of the product, the GTIN of each type is recorded.

  Promotion Data:
  - Promotion data contains the targets: price, regular price, and relative discount or absolute discount.

## Usage

You can load and use the dataset with the Hugging Face `datasets` library.
```python
import pandas as pd
from datasets import load_dataset

image_dataset = load_dataset("blamm/retail_visual_rag_pipeline", trust_remote_code=True)

product_promotion_data = load_dataset("blamm/retail_visual_rag_pipeline", data_files={'train':'train.parquet', 'test':'test.parquet'})
df_test = product_promotion_data['test'].to_pandas()

filename = '1166.jpg'
example_data = df_test.loc[df_test.filename == filename]
# Show product and promotion data
print(example_data)

image = image_dataset['test']['image'][example_data.index[0]]
# Show image
image.show()
```

<img src="figures/example_dataset_sample.png" alt="data_sample" width="300"/>

The *Dataset for Visual RAG pipeline* is used to evaluate the introduced Visual RAG pipeline. See the paper for explaination and evaluation of the Visual RAG pipeline.

## License
This dataset is licensed under a [Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International](https://creativecommons.org/licenses/by-nc-nd/4.0/)