---
tags:
  - deep-learning
  - pytorch
---

# Segmentation multiclass model 🚀

This model is a SegFormer5 trained on Sentinel2 images and CLCPlusBackBone Copernicus labels. The Sentinel2 images have 14 bands (original 12 bands + NDVI and NDWI) and their size is 250*250 pixels (resolution = 10m). 

## Description of the classes

- 1: Sealed
- 2: Conifers
- 3: Deciduous trees
- 4: Evergreen trees
- 5: Shrubs
- 6: Perennial herbaceous plants
- 7: Seasonal herbaceous plants
- 8: Lichens/mosses
- 9: Rocks, desert
- 10: Water
- 11: Snow and ice
- 0: Other

## Download the model
```python
from huggingface_hub import hf_hub_download

model_path = hf_hub_download("InseeFrLab/hackathon-ntts-2025-segmentation", "segmentation-multiclass-model-french-nsi-hackathon-ntts-2025.pth")
```