🔥🛰️ CanadaFireSat Model

July 5, 2026 · View on GitHub

This repository contains the code for training models on the benchmark CanadaFireSat available online. In this benchmark, we investigate the potential of deep learning with multiple sensors for high-resolution wildfire forecasting.

💿 Dataset on Hugging Face
📝 Published paper from ISPRS (ArXiv Version)
📊 Data repository on GitHub
🤖 Model Weights on Hugging Face

Disclaimer: [23/06/2026] This repository does not contain yet the models leveraging ignition proxy predictors used in Appendix E.

Summary Representation:

Model Architectures

In this repository, we train models following two different deep learning architectures, CNN-based using ResNet encoders and Transformer-based using ViT encoders.

Those models are trained across three data settings namely:

Setting	Source	Format	Type
SITS ONLY	Sentinel-2	Spatial	Multi-Spectral Images
ENV ONLY	MODIS	Spatial	Environmental Products
	ERA5-Land	Spatial	Climate Reanalysis
	CEMS	Spatial	Fire Indices
Multi-Modal	Sentinel-2	Spatial	Multi-Spectral Images
	MODIS	Tabular	Environmental Products
	ERA5-Land	Tabular	Climate Reanalysis
	CEMS	Tabular	Fire Indices

CNN-Based Multi-Modal Architecture

ViT-Based Multi-Modal Architecture

🛠️ Set-Up

⚠️ Support Python 3.10 (tested with 3.10.6).
In order to log model training, you need to set up a Weights & Biases (W&B) account or switch to a different logger. You can specify your W&B credentials in global_config.yaml.

1. Clone the repository

Clone the repository with its submodules:

git clone --recurse-submodules https://github.com/eceo-epfl/CanadaFireSat-Model.git
cd CanadaFireSat-Model

If you have already cloned the repository without submodules, initialize them with:

git submodule update --init --recursive

This will fetch the required DeepSatModels submodule located at:

deepsat/

2. Create the Python environment

python -m venv fire-env
source fire-env/bin/activate
pip install --upgrade pip setuptools wheel
pip install -r requirements/requirements.txt --extra-index-url https://download.pytorch.org/whl/cu117

This repository was tested with CUDA 11.7. If you use a different CUDA version or CPU-only installation, install the appropriate PyTorch build before installing the remaining dependencies.

3. Download the dataset

Download the data from Hugging Face 🤗 using src.huggingface.download with the configuration in download.yaml.

4. Configure paths

Specify the dataset and output paths in global_config.yaml before training.

🏋️ Model Training & Evaluation

Training: Run the src.train.segmentation_training script with your selected training config: ResNet_MULTI.yaml, ViT_MULTI.yaml, ...
Evaluation: Run the src.eval.eval script with your selected evaluation config: eval.yaml, eval_tab.yaml, ... . The model config described in the evaluation should match the one used for training.

📷 Results

📊 Performance Analysis: In this table, we describe the models' performances across data settings and architectures.

Encoder	Modality	Params (M)	Val		Test		Test Hard		Avg
Encoder	Modality	Params (M)	PRAUC	F1	PRAUC	F1	PRAUC	F1	PRAUC	F1
ResNet-50	SITS Only	52.2	45.2	49.3	53.3	58.9	26.3	36.7	41.6	48.3
	ENV Only	97.5	41.6	46.7	49.9	53.5	24.5	33.1	38.7	44.4
	Multi-Modal	52.2	46.1	51.1	57.0	60.3	27.1	37.4	43.4	49.6
ViT-S	SITS Only	36.5	45.2	50.6	51.2	51.9	25.7	33.8	40.7	45.2
	ENV Only	54.8	34.8	45.7	49.2	59.9	21.2	35.1	35.1	46.9
	Multi-Modal	37.7	43.9	50.0	56.3	59.2	25.1	36.6	41.8	48.6
Baseline (FWI)	ENV Only	-	20.0	32.7	43.1	50.3	21.1	32.7	28.1	38.6
Baseline (U-Net)¹	ENV Only	9.1	33.6	43.2	51.4	58.4	25.1	34.2	36.7	45.3
Baseline (U-TAE)²	ENV Only	1.1	32.9	43.8	47.2	52.5	22.0	31.7	34.0	42.7
Baseline (ConvLSTM)³	SITS Only	1.2	41.4	46.0	50.2	58.9	23.1	35.0	38.2	46.6

¹ Prapas et al., 2023 ² Michail et al., 2025 ³ Yang et al., 2021

🗺️ Use Cases on large ROI: We plot a large target area where a wildfire occurred in Québec in 2023, then the fire polygons corresponding to the wildfires, then our model predictions across the region.

Figure 1: Sentinel-2 tile from 2023/06/28 of size 14 km × 26 km before a large wildfire in Québec.

Figure 2: Fire polygons for the large wildfire on 2023/07/05 over the same tile.

Figure 3: Binary model predictions (in red) over the 2.64 km × 2.64 km center-cropped positive samples outlined in black.

🖋️ Citation

The paper has been published in the ISPRS Journal of Photogrammetry and Remote Sensing.

@article{porta2026canadafiresat,
  title={CanadaFireSat: Towards high-resolution wildfire forecasting with multiple modalities},
  author={Porta, Hugo and Dalsasso, Emanuele and McCarty, Jessica L and Tuia, Devis},
  journal={ISPRS Journal of Photogrammetry and Remote Sensing},
  volume={239},
  pages={555--572},
  year={2026},
  publisher={Elsevier}
}