Detection of fire risk patterns on the gas station by a trained neural network

The article analyzes the architecture of convolutional neural networks, which allows the detection of fire risk patterns without the use of significant computing power and expensive equipment. The process of training a neural network on its own data sets is also considered.