Machine Learning Techniques for Detection of Forest Fire

Forests are an essential natural resource for humanity because they provide a variety of direct and indirect advantages. It is essential to investigate the automatic detection of forest fires to prevent catastrophes. First, to address the issue of unequal distribution and a lack of samples, high-quality samples of forest fires were produced using General Advanced Networks, or GAN. Second, the forest fire area picture was predicted using an Adaboost classifier based on the properties of HOG (Histogram of Oriented Gradients), and the fire area was recognized secondarily using convolutional neural networks (CNN) and other methods: support vector machines, or SVMs. We want to automatically identify forest fires using deep learning (DL) based on the dataset of pictures (obtained by planes, satellites, and drones). In this study, we focused on the building of a deep learning model which will designed in such a way that it can detect wildfires using transfer learning methods from the best-trained deep learning computer vision architectures. Our proposed approach demonstrated its applicability in actual forest fire detection applications by achieving a detection rate of greater accuracy across various parameters.