Real Fundus Clinical Image Restoration Benchmark

One hundred twenty paired low quality and high quality clinical fundus photographs of the same eyes at 2560 by 2560 pixels for image restoration and enhancement research.

At a glance

Field	Value
Short name	real_fundus
Full name	Real Fundus Clinical Image Restoration Benchmark
Primary category	fundus
Contained modalities	fundus
Tasks	restoration
Samples	120
Classes	Not reported (Not reported)
Splits	all
Size	1.02 GB
Source-stated terms	Unknown; no dataset license declared in the repository
Normalized terms	unknown
Descriptive screening label	Unknown or unclear; do not assume permission
Terms scope	unknown
Access friction	anonymous_direct
Route backend	Manual (upstream-gated)
Availability	available (checked 2026-07-21)
Acquisition support	guided_instructions_only
Legacy sample-loader status	Metadata and access only

Notes

The public GitHub release contains the image archive. The paper describes a random 81/9/30 experimental split, but the release does not define distributed train, validation, and test partitions. No license file or dataset use terms were found.

Access preflight and acquisition

CLI

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

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

Python

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

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

Upstream page: github.com/dengzhuo-AI

Source-term evidence: github.com/dengzhuo-AI

Loader status

This catalog record provides metadata and access instructions, but it does not yet include a standard DatasetSample loader. Inspect the source file structure or contribute a loader before using it in a training pipeline.

Citation

BibTeX

@misc{real_fundus,
  title  = { Real Fundus Clinical Image Restoration Benchmark },
  note   = { Deng Z, Cai Y, Chen L, et al. RFormer: Transformer-based generative adversarial network for real fundus image restoration on a new clinical benchmark. IEEE J Biomed Health Inform. 2022;26:4645-4655. doi:10.1109/JBHI.2022.3187103 },
  year   = { 2022 },
  url    = { https://github.com/dengzhuo-AI/Real-Fundus/releases/tag/v.1.0.0 },
}

Plain text

Deng Z, Cai Y, Chen L, et al. RFormer: Transformer-based generative adversarial network for real fundus image restoration on a new clinical benchmark. IEEE J Biomed Health Inform. 2022;26:4645-4655. doi:10.1109/JBHI.2022.3187103

Source-stated terms

• Raw source string: Unknown; no dataset license declared in the repository
• Normalized category: unknown
• Apparent scope: unknown
• Descriptive screening label: Unknown or unclear; do not assume permission

⚠️ 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

• airogs: AIROGS: AI for Robust Glaucoma Screening (113,893 records, cc-by-nc-nd)
• eyecare_100k: Eyecare-100K: Multimodal Ophthalmology VQA Corpus (102,000 records, unknown)
• justraigs: JustRAIGS: Just Referral AI Glaucoma Screening Dataset (101,442 records, cc-by-nc-nd)
• eyepacs: EyePACS — Diabetic Retinopathy Detection (Kaggle 2015) (88,702 records, research-only)
• multieye: MultiEYE: OCT-Enhanced Fundus Multi-Disease Benchmark (58,036 records, mit)
• angioreport: AngioReport Fundus Angiography Report Dataset (55,361 records, unknown)
• ffa_ir: FFA-IR Medical Report Dataset (47,247 records, unknown)
• bidr: BiDR: Diabetic Retinopathy Diagnosis Dataset (35,126 records, unknown)