This project aims to identify suitable microsites for reforestration after disturbances such as cuttingh and/or fire. To acheive this goal aerial images are analyzed using machine learning alogrithms.
Following disturbances, reforestation is constrained by limited field access and the cost of manual site assessment and seeding.
This project trains a U-Net based semantic segmentation model to high-resolution images to automatically locate appropriate microsites to guide aerial seeding missions.
- Prepare and preprocess images for model training.
- Train and evaluate U-Net segmentation models for microsite detection.
- Generate georeferenced prediction layers for operational field planning.
- Track model improvements and document methodological decisions.
- docs
- examples
- papers
- src
- tests