JonathanFhima/RLAD-SwinV2-Large

Enhancing Retinal Vessel Segmentation Generalization via Layout-Aware Generative Modelling

State-of-the-art AV segmentation in retinal fundus images

Jonathan Fhima  •  Jan Van Eijgen  •  Lennert Beeckmans  •  Thomas Jacob  •  Moti Freiman
Luis Filipe Nakayama  •  Ingeborg Stalmans  •  Chaim Baskin  •  Joachim A. Behar

The strongest open-weight model for retinal artery/vein segmentation we know of, trained with RLAD-generated images and the REYIA dataset. Internally also known as LUNetV2.

Quickstart

The model loads the standard Hugging Face way - AutoModel and AutoImageProcessor with trust_remote_code=True download the code, config, weights, and preprocessing directly, nothing to fetch by hand:

import numpy as np
import torch
from PIL import Image
from transformers import AutoModel, AutoImageProcessor

repo = "JonathanFhima/RLAD-SwinV2-Large"
model = AutoModel.from_pretrained(repo, trust_remote_code=True).to("cuda").to(torch.bfloat16).eval()
processor = AutoImageProcessor.from_pretrained(repo, trust_remote_code=True)

image = Image.open("fundus.png").convert("RGB")
inputs = processor(image, return_tensors="pt")["pixel_values"].to("cuda", torch.bfloat16)
with torch.no_grad():
    output = model(inputs).float()

probs = torch.sigmoid(output)[0].cpu().numpy()
mask = np.zeros((*probs.shape[1:], 3), dtype=np.uint8)
mask[probs[0] > 0.5, 0] = 255                        # arteries: red
mask[(probs[1] > 0.5) & ~(probs[0] > 0.5), 2] = 255  # veins: blue

Or just run the CLI script, which does the same thing plus writes a blended overlay image:

python3 -m venv lunetv2_env && source lunetv2_env/bin/activate
pip install -r requirements.txt
python infer.py fundus.png out/

Writes out/mask.png and out/overlay.png. Runs in bfloat16, the precision the model was trained in.

For a full pipeline from a raw fundus image to vasculature biomarkers, see the RLAD + PVBM end-to-end tutorial, which feeds this model's output into PVBM.

Example

Results

If you plan to use this model or compare against it, we recommend scoring against our benchmark rather than re-running every baseline yourself. The scorer that produced these numbers lives in the REYIA repo (scripts/eval/dice_score.py).

Artery/vein Dice score, averaged across the same local, external, and out-of-distribution (OOD) test sets; ∆ rows show the improvement RLAD adds on top of each backbone, and † marks results with data leakage during training — this release is the SwinV2-large row.

Ablations

License

The code (LUNetv2.py, infer.py) is Apache 2.0. The weights (model.safetensors) are non-commercial, research use only — the training set mixes in subsets that are themselves non-commercial or restricted:

Source	Status
UZLF	KU Leuven dataset
GRAPE	CC0 1.0 — see REYIA
MAGRABIA	CC BY-NC 4.0 — see REYIA
MESSIDOR	redistribution and commercial use explicitly prohibited — see REYIA
PAPILA	figshare license field reads GPL 3.0+ — likely a depositor mis-click, not corrected as of this writing — see REYIA
ENRICH	non-commercial use only, KU Leuven data-sharing protocol — see REYIA

Citation

Fhima, J., Van Eijgen, J., Beeckmans, L., Jacob, T., Freiman, M., Nakayama, L.F., Stalmans, I., Baskin, C., Behar, J.A. Enhancing Retinal Vessel Segmentation Generalization via Layout-Aware Generative Modelling. arXiv:2503.01190, 2025. https://arxiv.org/abs/2503.01190