ResNetDETECT

Digital Object Identifier

Early Automatic Detection of Forest Fires Using Residual Learning Neural Networks

 

Objective

Develop an early automatic forest fire detection system using residual learning neural networks. The networks to be developed shall present an excellent ability in the detection of small smoke plumes in images gathered in the visible light wavelengths due to two main reasons:

  • The neural networks will be deep and their architecture will be chosen among several possible and optimized for the task in hands
  • The network training will be done using images created with highly advanced Generative Adversarial Networks as well as using a large number of real images gathered by CICLOPE system during several years.