---
license: mit
---

Model weights from the release at https://github.com/fieldsoftheworld/ftw-baselines/releases/tag/Pretrained-Models

The "full" models are trained on all country data in the Fields of The World dataset, while the CCBY models are trained according to the subset described [here](https://github.com/fieldsoftheworld/ftw-baselines?tab=readme-ov-file#cc-byor-equivalent-trained-models).


Example usage:
```
import segmentation_models_pytorch as smp
import torch

model = smp.Unet(
    encoder_name="efficientnet-b3",
    encoder_weights=None,
    in_channels=8,
    classes=2
)
model.load_state_dict(torch.load("ftw-2class-full_unet-efficientnetb3_rgbnir_f2444768.pth", weights_only=True))
```