FungalZSL: Zero-Shot Fungal Classification with Image Captioning Using a Synthetic Data Approach
Anju Rani, Daniel O. Arroyo, Petar Durdevic · 2025
Read FungalZSL: Zero-Shot Fungal Classification with Image Captioning Using a Synthetic Data Approach by Anju Rani, Daniel O. Arroyo, Petar Durdevic (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-2502-19038.txt.
Description
Naming the growth stage of a fungus from a photo is hard for image models when no labeled examples of that fungus exist. Anju Rani, Daniel Arroyo, and Petar Durdevic build FungalZSL, a zero-shot classifier that leans on synthetic data instead. A large language model writes text describing the stages of fungal growth, and synthetic images of fungi are created to go with the words.
Both the text and the images are projected into the shared space of a vision-language model, CLIP. That shared space lets pictures be matched to stage descriptions the model never saw paired in training. The team also compares text written by different language-model setups to see which wording sharpens the sorting. The aim is a classifier that can place new fungal images into growth stages without a costly hand-labeled photo set.
Catalog record
| CREATOR | Anju Rani, Daniel O. Arroyo, Petar Durdevic |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-19038