---
dataset_info:
  features:
  - name: image
    dtype: image
  splits:
  - name: train
    num_bytes: 3433643701.028
    num_examples: 48362
  download_size: 608599424
  dataset_size: 3433643701.028
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: mit
pretty_name: big-dataset-fundus-images
---

# Dataset Card for Dataset Name

<!-- Provide a quick summary of the dataset. -->

Fundus Images for Self Supervised Learning (FISSL)

## Dataset Details

### Dataset Description

<!-- Provide a longer summary of what this dataset is. -->

**FISSL** is a dataset resulting from the combination of 4 different retinal image datasets (RFMID, ODIR, eyePACS, APTOS). 
All images have been resized to 224x224 and converted to .png format

- **Curated by:** Diego Hernández
- **License:** MIT

### Dataset Sources

<!-- Provide the basic links for the dataset. -->

- **Repository:**
  - [huggingface dataset RFMID](https://huggingface.co/datasets/bumbledeep/rfmid)
  - [huggingface dataset ODIR](https://huggingface.co/datasets/bumbledeep/odir)
  - [huggingface dataset eyePACS](https://huggingface.co/datasets/bumbledeep/eyepacs)
  - [huggingface dataset APTOS](https://huggingface.co/datasets/bumbledeep/aptos)

## Uses

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

### Direct Use

<!-- This section describes suitable use cases for the dataset. -->

This version of the dataset is unlabeled and intended for feature extraction. More specifically, it was designed with **self-supervised learning** in mind.

## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

## Dataset Creation

### Curation Rationale

<!-- Motivation for the creation of this dataset. -->

The images have not been edited except for resizing. No additional processing has been performed.

### Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->

The used sources are properly described in each dataset card of the list given in the Dataset Sources section. 

#### Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

Composition of the dataset (original sources):

- [RFMID](https://www.kaggle.com/datasets/andrewmvd/retinal-disease-classification)
- [ODIR](https://www.kaggle.com/datasets/andrewmvd/ocular-disease-recognition-odir5k)
- [eyePACS](https://www.kaggle.com/datasets/tanlikesmath/diabetic-retinopathy-resized)
- [APTOS](https://www.kaggle.com/datasets/sovitrath/diabetic-retinopathy-224x224-2019-data/)

#### Who are the source data producers?

<!-- This section describes the people or systems who originally created the data. It should also include self-reported demographic or identity information for the source data creators if this information is available. -->

See the dataset cards of each source.


## Bias, Risks, and Limitations

<!-- This section is meant to convey both technical and sociotechnical limitations. -->

The heterogeneity in the images can be quite high due to the lack of a rigorous standardization process. This may affect the quality of the analyses performed.

### Recommendations

<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->

It is recommended to preprocess the images before using them to train deep learning models.

## Citation

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

If you use this dataset, please consider mentioning this repository or my Hugging Face username :)


## Dataset Card Authors

bumbledeep