Pyrovision: wildfire early detection

The increasing adoption of mobile phones have significantly shortened the time required for firefighting agents to be alerted of a starting wildfire. In less dense areas, limiting and minimizing this duration remains critical to preserve forest areas.

Pyrovision aims at providing the means to create a wildfire early detection system with state-of-the-art performances at minimal deployment costs.

Quick Tour

Automatic wildfire detection in PyTorch

You can use the library like any other python package to detect wildfires as follows:

from pyrovision.models.rexnet import rexnet1_0x
from torchvision import transforms
import torch
from PIL import Image


# Init
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])

tf = transforms.Compose([transforms.Resize(size=(448)), transforms.CenterCrop(size=448),
                         transforms.ToTensor(), normalize])

model = rexnet1_0x(pretrained=True).eval()

# Predict
im = tf(Image.open("path/to/your/image.jpg").convert('RGB'))

with torch.no_grad():
    pred = model(im.unsqueeze(0))
    is_wildfire = torch.sigmoid(pred).item() >= 0.5

Setup

Python 3.6 (or higher) and pip/conda are required to install Holocron.

Stable release

You can install the last stable release of the package using pypi as follows:

pip install pyrovision

or using conda:

conda install -c pyronear pyrovision

Developer installation

Alternatively, if you wish to use the latest features of the project that haven't made their way to a release yet, you can install the package from source:

git clone https://github.com/pyronear/pyro-vision.git
pip install -e pyro-vision/.

What else

Documentation

The full package documentation is available here for detailed specifications.

Docker container

If you wish to deploy containerized environments, a Dockerfile is provided for you build a docker image:

docker build . -t <YOUR_IMAGE_TAG>

Reference scripts

You are free to use any training script, but some are already provided for reference. In order to use them, install the specific requirements and check script options as follows:

pip install -r references/requirements.txt
python references/classification/train.py --help

You can then use the script to train tour model on one of our datasets:

Wildfire

Download Dataset from https://drive.google.com/file/d/1Y5IyBLA5xDMS1rBdVs-hsVNGQF3djaR1/view?usp=sharing

This dataset is protected by a password, please contact us at contact@pyronear.org

python train.py WildFireLght/ --model rexnet1_0x --lr 1e-3 -b 16 --epochs 20 --opt radam --sched onecycle --device 0

OpenFire

You can also use out opensource dataset without password

python train.py OpenFire/ --use-openfire --model rexnet1_0x --lr 1e-3 -b 16 --epochs 20 --opt radam --sched onecycle --device 0

You can use our dataset as follow:

from pyrovision.datasets import OpenFire
dataset = OpenFire('./data', download=True)

Citation

If you wish to cite this project, feel free to use this BibTeX reference:

@misc{pyrovision2019,
    title={Pyrovision: wildfire early detection},
    author={Pyronear contributors},
    year={2019},
    month={October},
    publisher = {GitHub},
    howpublished = {\url{https://github.com/pyronear/pyro-vision}}
}

Contributing

Please refer to CONTRIBUTING to help grow this project!

License

Distributed under the Apache 2 License. See LICENSE for more information.

MohammedElmehdiMehaya / pyro-vision