File size: 6,284 Bytes
560f1a0
 
499bdac
 
 
 
 
 
 
 
 
 
 
 
560f1a0
 
499bdac
560f1a0
499bdac
27835ad
499bdac
ace6f19
499bdac
27835ad
499bdac
fca5ad4
499bdac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
---
license: cc-by-nc-4.0
library_name: pytorch
pipeline_tag: image-segmentation
datasets:
- DailyWildfireCAN/DailyWildfireCAN
tags:
- wildfire
- wildfire-spread-prediction
- remote-sensing
- spatio-temporal
- benchmark
- canada
arxiv: 2611.14207
---

# 🍁 DailyWildfireCAN Benchmark: Model Checkpoints

[📄 Paper](https://arxiv.org/abs/2611.14207) · [🌐 Project page](https://dailywildfirecan.github.io) · [💻 Code](https://github.com/hagerradi/DailyWildFireCAN) · [🗂️ Dataset](https://huggingface.co/datasets/DailyWildfireCAN/DailyWildfireCAN)

This repository contains the trained checkpoints for the models in the **DailyWildfireCAN** benchmark for next-day wildfire spread prediction across Canada. Each model was trained three times with different random seeds, and all three runs are released.

![DailyWildfireCAN benchmark](https://cdn-uploads.huggingface.co/production/uploads/69f8c2efff7c500b6101eddd/kZlcdvFOCwqimxEmgMZFW.png)

## Task

Given the inputs available up to day *T*, predict which pixels will newly burn on day *T+1*.

- **Input:** a 256 × 256 tile at 90 m resolution with 50 channels (see the [dataset card](https://huggingface.co/datasets/DailyWildfireCAN/DailyWildfireCAN) for the channel order). Multi-temporal models take 3 consecutive days: (3, 50, 256, 256).
- **Output:** one logit per pixel, (1, 256, 256). Apply a sigmoid to get the probability of new burn.
- **Metrics:** AUC-PR, precision and recall, computed on the Test-Easy (unseen year), Test-Medium (unseen regions) and Test-Hard (unseen regions and year) splits.

## Models and results

AUC-PR is the mean over three seeds. Full results with precision, recall and standard deviations are in Tables 5 and 6 of the [paper](https://arxiv.org/abs/2611.14207).

### Mono-temporal models (single day of input)

| Model | Folder | `architecture` in code | Size / run | Test-Easy | Test-Medium | Test-Hard |
|---|---|---|---|---|---|---|
| U-Net | `unet` | `unet` | 373 MB | 0.285 | 0.218 | 0.175 |
| U-Net + Time-Gap Encoding | `unet_with_time_gap` | `unet_time_gap` | 374 MB | **0.287** | **0.233** | 0.192 |
| Attention U-Net | `attention_unet` | `unet_attention` | 377 MB | 0.280 | 0.199 | 0.155 |
| SegFormer U-Net | `segformer_unet` | `unet_segformer` | 571 MB | 0.205 | 0.181 | 0.162 |
| Swin-U-Net | `swin_unet` | `asufm` | 413 MB | 0.185 | 0.148 | 0.102 |
| OlmoEarth | `olmoearth` | `unet_olmo_offline` | 458 MB | 0.277 | 0.221 | 0.183 |
| DINOv3 | `dinov3` | `dual_dinov3` | 1.7 GB | 0.238 | 0.194 | **0.189** |

### Multi-temporal models (3 consecutive days of input)

| Model | Folder | `architecture` in code | Size / run | Test-Easy | Test-Medium | Test-Hard |
|---|---|---|---|---|---|---|
| U-TAE | `utae` | `utae` | 139 MB | **0.352** | **0.306** | **0.259** |
| U-Net ConvLSTM | `unet_with_convlstm` | `unet_convlstm` | 348 MB | 0.340 | 0.278 | 0.222 |
| SimVPv2 | `simvpv2` | `simvpv2_spatiotemporal` | 66 MB | 0.284 | 0.250 | 0.203 |
| Video Swin-U-Net | `video_swin_unet` | `video_swin_unet` | 539 MB | 0.272 | 0.235 | 0.178 |
| Multi-Modal | `multi_mode_model` | `frp` | 4 MB | 0.182 | 0.167 | 0.110 |

Model descriptions and references are in the [code repository](https://github.com/hagerradi/DailyWildFireCAN#-part-4-modeling).

## Repository structure

```
benchmark_models_checkpoints/
├── unet/
│   ├── checkpoints_unet_run_1/best_checkpoint.pt
│   ├── checkpoints_unet_run_2/best_checkpoint.pt
│   └── checkpoints_unet_run_3/best_checkpoint.pt
├── unet_with_time_gap/
│   └── ...
└── <model>/checkpoints_<model>_run_<N>/best_checkpoint.pt
```

The full repository is about 16 GB. `best_checkpoint.pt` is the checkpoint with the best validation score. It is a dictionary with the keys `model_state_dict`, `optimizer_state_dict`, `scheduler_state_dict`, `epoch` and `best_val_loss`.

## Quick start

Clone the code repository, which contains the model definitions:

```bash
git clone https://github.com/hagerradi/DailyWildFireCAN
cd DailyWildFireCAN
pip install -r requirements.txt
```

Download a checkpoint and run inference:

```python
import torch
from huggingface_hub import hf_hub_download
from src.models.unet import UNet

path = hf_hub_download(
    repo_id="DailyWildfireCAN/DailyWildfireCAN-Benchmark",
    filename="benchmark_models_checkpoints/unet/checkpoints_unet_run_1/best_checkpoint.pt",
)

model = UNet(input_channels=50, num_classes=1, hidden_features=[64, 128, 256, 512])
ckpt = torch.load(path, map_location="cpu")
model.load_state_dict(ckpt["model_state_dict"])
model.eval()

x = torch.randn(1, 50, 256, 256)        # replace with a normalized sample's "x"
with torch.no_grad():
    prob = torch.sigmoid(model(x))      # (1, 1, 256, 256) probability of new burn
```

For the other models, build the model the same way as in `src/evaluate.py` with the settings in `configs/default.yaml`, then load `model_state_dict`.

Download all runs of one model:

```python
from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="DailyWildfireCAN/DailyWildfireCAN-Benchmark",
    allow_patterns=["benchmark_models_checkpoints/utae/*"],
    local_dir="checkpoints",
)
```

## Training details

All models were trained with the same setup, using the full set of input features and the training-set normalization statistics released with the dataset. See Appendix B of the [paper](https://arxiv.org/abs/2611.14207) for hyperparameters and hardware.

## License

The checkpoints are released under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. The models were trained on [DailyWildfireCAN](https://huggingface.co/datasets/DailyWildfireCAN/DailyWildfireCAN), which is released under CC BY-NC-SA 4.0 and includes third-party data subject to their own licenses.

## Citation

```bibtex
@article{dailywildfirecan,
  title={DailyWildfireCAN: A Dataset and Benchmark for Next-Day Wildfire Spread Prediction in Canada},
  author={Ettahri, Mahacine and Abdelwahed, Hager Radi and Olutayo, Toluwaleke and Aloise, Daniel and Caron, Ga{\'e}tan Marceau},
  journal={arXiv preprint arXiv:2611.14207},
  year={2026}
}
```

## Contact

For inquiries, contact Mahacine Ettahri (mahacine.ettahri@etud.polymtl.ca) or Hager Radi (hager.radi@mila.quebec).