UWF DR Reasoning Dataset

Ultra-widefield fundus photographs for diabetic retinopathy grading (5-level ICDR scale). Released alongside the Reasoning-Enhanced VLM paper for interpretable UWF DR detection. Related to the UWF4DR 2024 MICCAI challenge (~495 images across image quality and DR grading tasks).

At a glance

Field	Value
Short name	uwf_dr
Full name	UWF DR Reasoning Dataset
Primary category	uwf_fundus
Contained modalities	uwf_fundus
Tasks	grading, classification
Samples	495
Classes	5 (No DR, Mild DR, Moderate DR, Severe DR, Proliferative DR)
Splits	train, val, test
Size	2.0 GB
Source-stated terms	Research only — see paper terms
Normalized terms	research-only
Descriptive screening label	Research or challenge restriction recorded; check source
Terms scope	dataset_files
Access friction	anonymous_direct
Route backend	Google Drive
Availability	available (checked 2026-07-21)
Acquisition support	guided_instructions_only
Legacy sample-loader status	Standard loader included

Access preflight and acquisition

CLI

# Read-only preflight
eyehub download uwf_dr --data-dir ./data --dry-run --json

# Explicit transfer, only when preflight reports supported behavior
eyehub download uwf_dr --data-dir ./data

Python

from eyedatahub.acquisition import preflight_dataset
from eyedatahub.datasets.registry import REGISTRY

ds = REGISTRY.get_dataset('uwf_dr')
print(preflight_dataset(ds, './data'))  # no transfer

Upstream page: drive.google.com/drive

Source-term evidence: drive.google.com/drive

Loader example

This entry includes a standard DatasetSample loader.

from pathlib import Path
from eyedatahub.datasets.registry import REGISTRY

data_dir = Path('~/.eyedatahub/data').expanduser()
ds = REGISTRY.get_dataset('uwf_dr')
samples = ds.load(data_dir, split='test')
for s in samples[:5]:
    print(s.sample_id, s.label, s.image_path)

Citation

BibTeX

@misc{uwf_dr,
  title  = { UWF DR Reasoning Dataset },
  note   = { Reasoning-Enhanced Vision-Language Model for Interpretable Diabetic Retinopathy Detection in Ultra-Wide-Field Fundus Images. OMIA 2025 (MICCAI Workshop), Springer. DOI: 10.1007/978-3-032-10351-2_12 },
  year   = { 2025 },
  url    = { https://drive.google.com/drive/folders/1wOqM-O_amSwMdli4OlFnGpf4hslPgDNu },
}

Plain text

Reasoning-Enhanced Vision-Language Model for Interpretable Diabetic Retinopathy Detection in Ultra-Wide-Field Fundus Images. OMIA 2025 (MICCAI Workshop), Springer. DOI: 10.1007/978-3-032-10351-2_12

Source-stated terms

• Raw source string: Research only — see paper terms
• Normalized category: research-only
• Apparent scope: dataset_files
• Descriptive screening label: Research or challenge restriction recorded; check source

⚠️ Source-stated terms, scope, and normalized labels are curation metadata, not legal advice or a permission finding. Review the current official source before transfer or reuse.

Related datasets with shared modalities

• mmrdr: MMRDR: Multi-Modal Retinal Diabetic Retinopathy Dataset (24,460 records, cc-by)
• oculoscope: OculoScope: Fairer AI in Ophthalmology Dataset (16,530 records, cc-by)
• tsukazaki_uwf: Tsukazaki Hospital UWF Fundus Dataset (13,047 records, research-only)
• birdshot_wide: Birdshot-WIDE Widefield Fundus Dataset (6,352 records, cc-by)
• deepdrid: DeepDRiD: Diabetic Retinopathy Grading and Image Quality Dataset (2,256 records, cc-by-sa)
• uwf_tumor: UWF Fundus Intraocular Tumor Dataset (2,031 records, cc-by)
• uwf_dr_peng: UWF Fundus DR Dataset (Peng et al., 2026) (1,630 records, cc-by)
• uwf_zhejiang: Open UWF Fundus Dataset with Disease + Quality Labels (700 records, cc-by)