README.md

---
license: apache-2.0
library_name: pytorch
tags:
- methane-detection
- hyperspectral
- semantic-segmentation
- remote-sensing
- fourier-neural-operator
- aviris-ng
- emit
---

# FLAME pretrained weights

Weights for [FLAME: Physics-Guided Neural Operators for Onboard Satellite
Methane Detection in Hyperspectral Imagery](https://arxiv.org/abs/2606.01577).

Code: https://github.com/ROKMC1250/FLAME

| File | Trained on | Config |
|---|---|---|
| `flame_starcop.pt` | STARCOP | `flame_starcop.yaml` |
| `flame_emit.pt` | OxHyperSyntheticCH4 | `flame_emit.yaml` |

Each checkpoint is a `torch.save` dict with keys `model`, `epoch` and
`metric`. The matching training configs are included here and in the code
repository under `configs/`.

## Usage

```bash
pip install huggingface_hub
hf download hjh1037/FLAME flame_starcop.pt flame_emit.pt --local-dir .
```

See the [code repository](https://github.com/ROKMC1250/FLAME) for evaluation
and visualization instructions.

```python
import torch, yaml
from flame.model import build_model

cfg = yaml.safe_load(open('configs/flame_starcop.yaml'))
model = build_model(cfg['model']).eval()
state = torch.load('flame_starcop.pt', map_location='cpu', weights_only=False)
model.load_state_dict(state['model'])
```

## Citation

```bibtex
@article{heo2026flame,
  title={FLAME: Physics-Guided Neural Operators for Onboard Satellite Methane Detection in Hyperspectral Imagery},
  author={Heo, Junhyuk and Park, Junhwan and Sim, Sancheol and Choi, Beomkyu and Cho, Woojin},
  journal={arXiv preprint arXiv:2606.01577},
  year={2026}
}
```