Wildfire_detection

Detection of wildfire using information from unmanned drones.

more examples

Requirements:

• Pillow
• cython
• pytorch
• torchvision

Install all requireents in conda enviroment

conda env create -n <env-name> -f ./Detector/environment.yml

Dataset

Wildfire Smoke Dataset - contains 737 images presplited on train, test and validation samples + bounding boxes (RetinaNet Keras format is used). My version of dataset could be found on my drive.

Expanding dataset is planned in the future.

Installing pycocotools

not necessary (already stored in in ./lib/)

git clone https://github.com/cocodataset/cocoapi.git
cd cocoapi/PythonAPI
make
cp -r pycocotools/ ~/wildfire_detection/Detector/lib

Pretrained model

You can download serialised pretrainde model from my drive because it is too large for git.