Beyond Disease Labels: Symptom-Based Vision-Language Learning for Unseen Plant Diseases
Preprint, code and symptom annotations for learning to read visual plant-disease symptoms with vision-language models and evaluating on diseases withheld from training (10-fold unit-grouped cross-validation over 120 disease units, plus an external farmer-photo test). Code and project page: https://github.com/AmirrrezaZ/plant-diagnose-anything , https://amirrrezaz.github.io/plant-diagnose-anything/ . Code is MIT; annotations in data/ are CC BY 4.0. Photographs are not included (PlantWild, CC BY-NC-ND).
Authors
- Amirreza Zibafar
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-07
- DOI
- https://doi.org/10.5281/zenodo.24032334
- Primary Topic
- Smart Agriculture and AI
- Type
- preprint