File size: 2,661 Bytes
902ac20
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
---
license: other
language:
  - en
tags:
  - medical
  - image-segmentation
  - glaucoma
  - retinal-fundus
  - ophthalmology
task_categories:
  - image-classification
  - image-segmentation
pretty_name: REFUGE2
size_categories:
  - 1K<n<10K
---

# REFUGE2

Retinal fundus image dataset from the **REFUGE2 Challenge** (Retinal Fundus Glaucoma
Challenge, 2nd edition, MICCAI 2020 / OMIA workshop) — a standard benchmark for
evaluating automated algorithms for glaucoma detection and optic disc/cup
segmentation on retinal fundus photographs.

This repository mirrors the Kaggle dataset
[victorlemosml/refuge2](https://www.kaggle.com/datasets/victorlemosml/refuge2)
(4.2 GB, 2,400 files), uploaded for convenient access via Hugging Face.

## Challenge tasks

1. **Classification** of clinical glaucoma
2. **Segmentation** of the optic disc and cup
3. **Localization** of the fovea (macular center)

## Dataset structure

| Split | Images | Masks | Notes |
|---|---|---|---|
| train | 400 JPG (`g####.jpg`) | 400 BMP | glaucoma cases prefixed `g`, non-glaucoma `n` |
| val | 400 JPG (`V####.jpg`) | 400 PNG | |
| test | 400 JPG (`T####.jpg`) | 400 BMP | |

```
train/images/*.jpg   train/mask/*.bmp
val/images/*.jpg     val/mask/*.png
test/images/*.jpg    test/mask/*.bmp
```

- **Images**: color fundus photographs (JPEG), centered at the posterior pole with
  both the macula and optic disc visible. The REFUGE data were acquired with Zeiss
  Visucam 500 (2124×2056) and Canon CR-2 (1634×1634) cameras; REFUGE2 added images
  from Kowa (validation) and Topcon (test) devices.
- **Masks**: pixel-wise optic disc/cup reference segmentations produced by seven
  independent glaucoma specialists and merged by a senior specialist.
  Pixel values: `0` = optic cup, `128` = optic disc, `255` = background.
- **Glaucoma labels**: reference standard from full clinical records (follow-up
  fundus images, IOP, OCT, visual fields), not from the fundus image alone.
  About 10% of cases are glaucomatous.

## Original sources

- Challenge website: https://refuge.grand-challenge.org/Home2020/
- Kaggle mirror (source of this upload): https://www.kaggle.com/datasets/victorlemosml/refuge2
- Paper: Orlando et al., "REFUGE Challenge: A unified framework for evaluating
  automated methods for glaucoma assessment from fundus photographs",
  *Medical Image Analysis* 59, 101570 (2020).

## License / terms

The Kaggle listing marks the license as "Unknown". The REFUGE/REFUGE2 challenge
data are distributed by the challenge organizers for research purposes; please
refer to the terms on the challenge website before use and do not redistribute
for commercial purposes.