README.md

---
metrics:
- accuracy
- f1
pipeline_tag: image-segmentation
tags:
- geospatial
- forest
- species
- productivity
- age
---
# Models for "_Characterizing long-term boreal forest ecosystems dynamics in Khanty-Mansi Autonomous Okrug (Western Siberia) using time series remote sensing data and deep learning methods_"

## Paper

Paper is published at Remote Sensing Applications: Society and Environment and is available at (link here)

## Other data

Jupyter notebooks, statistics and model logs are published on [GitHub](https://github.com/simonreise/forest-segmentation-2025), maps are published on [Zenodo](https://doi.org/10.5281/zenodo.18054578)

## Models description

This repository contains SegFormer model checkpoints of 5 models that model dominant tree species, forest productivity class, forest type, age and growing stock in boreal forests of West Siberia.

The checkpoints are created with Remote Sensing Processor library, but can be loaded with Lightning too.