Prototype-Guided Diffusion: Visual Conditioning without External Memory
Bilal FAYE, Hanane AZZAG, Mustapha Lebbahe-mail: [email protected], [email protected], [email protected] · 2025
Read Prototype-Guided Diffusion: Visual Conditioning without External Memory by Bilal FAYE, Hanane AZZAG, Mustapha Lebbahe-mail: [email protected], [email protected], [email protected] (2025) 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-2508-09922.txt.
Description
Bilal Faye, Hanane Azzag, and Mustapha Lebbah present a new way to guide image-making models, in a paper from 2025. These models build pictures by clearing noise step by step, which costs time and computing power. Some rivals fetch sample images from a stored bank to guide the work. That bank takes space and adds rigid setup steps.
The team's model learns a small set of visual prototypes instead. It picks these patterns from clean image features during training. While a picture forms, the model lines up the noisy draft with the closest patterns. Tests show picture quality stays high while compute and storage costs fall. The method scales without a memory bank to maintain. It offers a lean path for fast, guided image generation.
Catalog record
| CREATOR | Bilal FAYE, Hanane AZZAG, Mustapha Lebbahe-mail: [email protected], [email protected], [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-09922