README.md

---
dataset_info:
  features:
  - name: Description
    dtype: string
  - name: Label
    dtype: string
  splits:
  - name: original
    num_bytes: 8415
    num_examples: 50
  - name: augmented
    num_bytes: 165260
    num_examples: 1000
  download_size: 13985
  dataset_size: 173675
configs:
- config_name: default
  data_files:
  - split: original
    path: data/original-*
  - split: augmented
    path: data/augmented-*
---

# Dataset Card for aedupuga/cards-image-dataset

### Dataset Description

This Dataset consists of food descriptions of different food and beverage items labelled with **Ingredient**, **Dish** or **Beverage**. It was created as part of a practice in generatign and augmenting text datasets.

- **Curated by:** Anuhya Edupuganti


## Uses

<!-- Address questions around how the dataset is intended to be used. -->

### Direct Use

- Training and evaluating text classification models 
- Experimenting with preporcessing of text datasets.

## Dataset Structure

This data set contains teo splits:
- **original**: 50 samples of handwritten descriptions of differnt food items.
- **augmented**: 300 examples generated by augmentation to expand data diversity and variety.

Each row includes:  

- `Description` (string): text description of the food.
- `label`(string): string class label 'Dish', 'Ingredient' and 'Beverage'



### Curation Rationale

This dataset was generated as a practice in text dataset creation and augmentation.


#### Data Collection and Processing

- Original data handwritten.
- Augmentation with backtranslation technique.


## Bias, Risks, and Limitations

- **Small sample size:** Only 50 original samples.  
- **Synthetic augmentation:** May not accurately describe food realisticly.

### Recommendations

- Use primarily to practice classification methods.

## Dataset Card Contact

Anuhya Edupuganti (Carnegie Mellon Univerity)- aedupuga@andrew.cmu.edu