A Novel Lightweight CNN Model for Real-Time Video Fire Smoke Detection

Surveillance has a variety of applications, including login detection, crime detection, and event monitoring. Many unusual events, such as fires, accidents, and disasters, can occur while under surveillance. Among such rare occurrences, fire is one of the most common causes. This requirement Three are the most effective fire alarm systems available for monitoring. Existing flame detection systems are based on visible light emitted by surveillance cameras. These methods, however, do not work well in small fires, and color-base The offered modes area unit divided into 3 types: pixel level, blob level, and patch level modes and are sensitive to fires, light, and shade. Convolutional neural networks have enabled vision-based systems to observe hearth throughout police investigation because of the most recent development of embedded process (CNNs). Such methods, on the other hand, often require a lot of time for processing and memory, limiting their use in monitoring networks. This research work presents the CNN's inexpensive fire detection architecture for security surveys. The model is based on the architecture of GoogleNet, which has a lower computing complexity and is better suited to challenge than other expensive networks such as Alex Net.The model is well-designed to balance effectiveness and delicacy, taking into account the type of target problem and fire data.