Deep Learning Model for Detection and Recognition of Fire based on Virtual Reality Video Images

Abstract

Fire detection and recognition is an important aspect of fire safety, and the use of virtual reality video images and deep learning (DL) methods can help to optimize this process. Deep learning (DL) is the sub-field of machine learning (ML) which utilizes the artificial neural networks (ANN) to train and analyze predictions. These networks are more suitable for processing enormous amounts of data which is better for image recognition. Based on the fire status and immersive view, the detection and recognition of fire are detected. Deep learning algorithms can be trained using these images to recognize patterns and identify fires, smoke, and other indicators of fire. This paper introduced the new fire detection model which detects the fire from video footage and also images collected various online sources. The proposed model used the pre-trained model RESNET-50 to train the fire affected videos. To detect the fire affected region the feature extraction method Histogram of Oriented Gradients and Radial Basis Function Networks (RBFNs) used to detect the fire affected images.