koyu008/RFMID

---
language:
- en
license:
- cc-by-4.0
tags:
- medical
- ophthalmology
- fundus-image
- image-classification
- multi-label
- diabetic-retinopathy
- glaucoma
- amd
- multi-disease-screening
task_categories:
- image-classification
task_ids:
- multi-label-image-classification
pretty_name: RFMiD (Retinal Fundus Multi-Disease Image Dataset)
size_categories:
- 1K<n<10K
annotations_creators:
- expert-generated
source_datasets:
- original
source_data_urls:
- https://www.mdpi.com/2306-5729/6/2/14
- https://ieee-dataport.org/open-access/retinal-fundus-multi-disease-image-dataset-rfmid
---

# 🩺 RFMiD — Retinal Fundus Multi-Disease Image Dataset

<table align="center">
<tr>
<td width="100%" align="center">
<img src="rm_images/Merged_Fundus_Images_with_Caption.jpg" alt="Merged Dataset Samples" style="max-width: 100%; height: auto;">
<br>
<p><strong>Image:</strong> Dataset Samples.</p>
</td>
</tr>
</table>

> The **Retinal Fundus Multi-Disease Image Dataset (RFMiD)** is designed for **multi-disease detection and classification** in retinal fundus photographs.
> It includes 3,200 high-quality color images with **46 labeled retinal disease conditions**, curated by expert ophthalmologists from India.
> This dataset enables development of generalized deep learning models for comprehensive retinal disease screening.

---

## 📘 Overview

| Field | Details |
|-------|----------|
| **Full Name** | Retinal Fundus Multi-Disease Image Dataset (RFMiD) |
| **Focus** | Multi-label classification of retinal diseases |
| **Condition Types** | 46 disease classes including diabetic retinopathy, glaucoma, AMD, hypertensive retinopathy, myopia, and others |
| **Collection Site** | Ophthalmology centers in Maharashtra, India |
| **Devices Used** | TOPCON 3D OCT-2000 (~2144×1424), Kowa VX-10α (~4288×2848), TOPCON TRC-NW300 (~2048×1536) |
| **Field of View (FOV)** | ~45°–50° |
| **Image Type** | Color fundus photographs (JPG, RGB) |
| **Total Images** | **3,200** |
| **Annotations** | Expert ophthalmologist-verified, multi-label (each image may contain multiple conditions) |
| **License** | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) |
| **Source** | [MDPI Paper](https://www.mdpi.com/2306-5729/6/2/14) · [IEEE Dataport](https://ieee-dataport.org/open-access/retinal-fundus-multi-disease-image-dataset-rfmid) |

---

## 🗂️ Dataset Structure

The RFMiD dataset includes **images and corresponding metadata** files organized as follows:

```text
RFMiD/
│
├── Images/
│ ├── Training_Set/
│ │ ├── IDRiD_001.jpg
│ │ ├── IDRiD_002.jpg
│ │ └── ...
│ │
│ ├── Validation_Set/
│ │ ├── IDRiD_801.jpg
│ │ ├── IDRiD_802.jpg
│ │ └── ...
│ │
│ └── Test_Set/
│ ├── IDRiD_901.jpg
│ ├── IDRiD_902.jpg
│ └── ...
│
├── Groundtruths/
│ ├── RFMiD_Training_Labels.csv
│ ├── RFMiD_Validation_Labels.csv
│ └── RFMiD_Test_Labels.csv
│
└── Metadata/
└── RFMiD_Clinical_Information.csv
```

---

### 📄 File Description

| File / Folder | Description |
|----------------|-------------|
| **Images/** | Contains all RGB fundus images grouped into train, validation, and test sets |
| **Groundtruths/** | CSV files with disease labels for each image ID |
| **Metadata/** | Contains additional information like patient age, gender, and diagnostic notes (if available) |

### 🧾 Label Format (CSV Example)

Each row in `RFMiD_Training_Labels.csv` includes binary indicators (0 or 1) for each of the 46 disease categories:

| ImageID | DR | ARMD | MH | DN | MYA | ... | HR | Others |
|----------|----|------|----|----|-----|-----|----|--------|
| 0001 | 1 | 0 | 0 | 1 | 0 | ... | 0 | 0 |
| 0002 | 0 | 0 | 0 | 0 | 0 | ... | 1 | 0 |

*Total columns: 46 disease labels + 1 ImageID column.*

---

## 📊 Dataset Composition

| Split | Number of Images | Description |
|--------|------------------|--------------|
| **Training Set** | 1,920 | Used to train AI models |
| **Validation Set** | 640 | Used to tune hyperparameters |
| **Test Set** | 640 | Held-out evaluation set |
| **Total** | **3,200** | All high-quality fundus images |

---

## 🧠 Research Applications

### Primary Use Cases
- Multi-label retinal disease classification
- Generalized ophthalmic AI screening
- Rare disease detection (long-tail recognition)
- Domain adaptation across imaging devices
- Quality-aware retinal analysis

### Recommended Tasks
- **Classification:** Healthy vs Abnormal
- **Multi-label Detection:** 46 retinal diseases
- **Transfer Learning:** Adaptation to real-world clinical data
- **Explainability:** Visualizing disease localization with Grad-CAM or attention maps

---

## ⚙️ Technical Notes

- **Input format:** RGB fundus images, JPG
- **Recommended preprocessing:** Center-cropping, illumination correction, resizing to 512×512 or 1024×1024
- **Label imbalance:** Some diseases have <50 samples; use focal loss or weighted sampling
- **Multi-device domain variation:** Apply histogram equalization or color normalization

---

## 🧩 Quick Summary Table

| Dataset | Description (conditions, source, etc.) | Size |
|----------|----------------------------------------|------|
| **RFMiD** | Multi-disease retinal fundus dataset with 46 labeled conditions from Indian ophthalmic clinics | **3,200 images** |

---

## 📚 Citation

If you use this dataset, please cite:

> **Pachade, S.; Porwal, P.; Thulkar, D.; Kokare, M.; Deshmukh, G.; Sahasrabuddhe, V.; Giancardo, L.; Quellec, G.; Mériaudeau, F.**
> *Retinal Fundus Multi-Disease Image Dataset (RFMiD): A Dataset for Multi-Disease Detection Research.*
> **Data** 2021, 6(2), 14.
> DOI: [10.3390/data6020014](https://doi.org/10.3390/data6020014)

---

## 🪪 License

This dataset is released under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** license.
You may share and adapt the dataset, provided appropriate credit is given.

---