/ForestFires

Forest Fire Danger Detection model

Primary LanguageJupyter NotebookMIT LicenseMIT

Wildfire danger optimization using Deep Learning and Transfer Learning

This is the code for every implementation for the paper: Wildfire danger prediction optimization using deep learning. We used Transfer Learning to create accurate CNN models to detect if a forest has been destroyed by a wildfire or not to help the already existing FWI prediction. We also created some Regression models trained on Greek Wildfires in 2018.We showed that there's a very low correlation between the meteorological data and the destroyed area,thus,almost every wildfire in Greece is because of human hands.

Tensorflow Pretrained CNN models

We used:

  • Xception
  • ResNet50
  • VGG16
  • VGG19
  • ResNetV2
  • EfficientNetV2L
  • EfficientNetB7
  • Some Custom CNN's based on the best results from above.