Gen4Regen Dataset Uses AI to Map Forest Regeneration Species
Researchers have introduced the Gen4Regen dataset, leveraging the Nano Banana Pro vision-language model to generate synthetic training data for semantic segmentation of forest regeneration species from UAV imagery. The work addresses severe scarcity of expert-annotated images and class imbalance in fine-grained species mapping. The framework produces high-fidelity images and pixel-aligned masks from prompts, reducing reliance on manual photo-interpretation. The dataset is part of the WilDReF-Q-V2 project, aimed at scalable forest management.
Key facts
- Gen4Regen dataset addresses data scarcity for forest regeneration mapping.
- Uses Nano Banana Pro model to generate synthetic images and masks.
- Focuses on semantic segmentation of fine-grained species.
- Aims to reduce manual photo-interpretation for UAV imagery.
- Part of WilDReF-Q-V2 project.
- Published on arXiv with ID 2605.05627.
Entities
Institutions
- arXiv