Text-to-Image Transformed MNIST Dataset (text2image_dataset)

Overview

This DeepTrackAI repository contains a dataset of transformed MNIST digits paired with natural-language descriptions.
Images are generated using the notebook create_transformed_images.ipynb, which applies randomized color, rotation/flip, and contrast transformations to the original MNIST digits and saves both the transformed images and corresponding text prompts.

The dataset is designed for benchmarking text-to-image alignment and multimodal learning tasks.

Summary

• Number of images: equal to the number of MNIST images (60,000), each with one transformed version
• Image size: 28 × 28 pixels
• Image format: 8-bit RGB PNG
• Metadata: JSON file mapping filenames to natural-language transformation descriptions
• Transformations included:
  • Digit colorization (e.g., “in red on a blue background”)
  • Rotations (±90°, 180°)
  • Horizontal mirror / vertical flip
  • Auto-contrast normalization

---

Original Source

• Base dataset

  • Title: The MNIST Database of Handwritten Digits
  • Authors: Yann LeCun, Corinna Cortes, Christopher J.C. Burges
  • Source: Official MNIST Website
  • Reference article: LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. Proceedings of the IEEE 86(11): 2278-2324 (1998). DOI: 10.1109/5.726791
  • License: Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0)
  • Note: The original MNIST website is no longer actively maintained. An archived version is available, and the dataset is widely mirrored (e.g., on Kaggle).

• Derivative dataset

  • Title: Text-to-Image Transformed MNIST Dataset
  • Authors: Carlo Manzo
  • Source: This repository
  • License: Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0)

If you use this dataset in your research, please follow the licensing requirements and properly attribute the original authors.

---

Dataset Structure

/text2image_dataset  
└── train/  
    ├── images/                 # Transformed digit images 
    │   ├── 0_000000.png  
    │   ├── 0_000001.png  
    │   └── ...  
    └── image_descriptions.json # Mapping of filenames to natural-language transformation descriptions  

Each filename inimages/ begins with the digit label (0–9), followed by a sequential numerical identifier. The JSON file provides the corresponding text description for each transformed image.

---

How to Access the Data

Clone the Repository

git clone https://github.com/DeepTrackAI/text2image_dataset  
cd text2image_dataset  

Generate the Dataset

Run the notebook to generate transformed images and their descriptions:

jupyter notebook create_transformed_images.ipynb

---

Attribution

When using this dataset or the code, please cite the original MNIST dataset, the reference article, and the text-to-image transformation repository.

Cite the dataset:

LeCun Y, Cortes C, Burges CJC. The MNIST Database of Handwritten Digits. Retrieved from http://yann.lecun.com/exdb/mnist/

@misc{lecun1998mnist,
  title        = {The MNIST Database of Handwritten Digits},
  author       = {LeCun, Yann and Cortes, Corinna and Burges, Christopher J.C.},
  year         = {1998},
  howpublished = {\url{http://yann.lecun.com/exdb/mnist/}}
}

Cite the reference article:

LeCun Y, Bottou L, Bengio Y, Haffner P. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11): 2278–2324 (1998). DOI: 10.1109/5.726791

@article{lecun1998gradient,
  title     = {Gradient-based learning applied to document recognition},
  author    = {LeCun, Yann and Bottou, L{\'e}on and Bengio, Yoshua and Haffner, Patrick},
  journal   = {Proceedings of the IEEE},
  volume    = {86},
  number    = {11},
  pages     = {2278--2324},
  year      = {1998},
  publisher = {IEEE},
  doi       = {10.1109/5.726791}
}

Cite this repository:

Carlo Manzo. Text-to-Image Transformed MNIST Dataset. GitHub (2025). GitHub

@misc{text2image2025,  
  author       = {Carlo Manzo},  
  title        = {Text-to-Image Transformed MNIST Dataset},  
  year         = {2025},  
  howpublished = {\url{https://github.com/DeepTrackAI/text2image_dataset}}  
}  

---

License

As a derivative dataset, this repository is shared under the Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) License, following the original licensing terms.