Ordinal Diffusion Models for Color Fundus Images
Gustav Schmidt, Philipp Berens, Sarah Müller · 2026
Read Ordinal Diffusion Models for Color Fundus Images by Gustav Schmidt, Philipp Berens, Sarah Müller (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-24013.txt.
Description
Gustav Schmidt and two co-authors present an ordinal diffusion model for generating color fundus images that respects the ordered stages of diabetic retinopathy. The 2026 paper notes that most conditional diffusion models treat disease stages as unrelated classes, ignoring that retinopathy progresses continuously.
Instead of categorical conditioning, the model uses a scalar disease representation, enabling smooth transitions between neighboring stages. Tests on a large eye-image dataset use visual realism metrics and classification-based clinical consistency checks. Compared with a standard conditional model, the ordinal model improves image realism for most disease stages and raises agreement between generated images and their intended stage. Put plainly, the model treats the disease as one slow shift, not a set of boxes. That view matches how the disease moves in real eyes.
Catalog record
| CREATOR | Gustav Schmidt, Philipp Berens, Sarah Müller |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-24013