Desenvolvimento de um software identificador de incêndios utilizando um modelo de machine learning
Nenhuma Miniatura disponível
Data
2023-08-14
Autores
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Goiás
Resumo
This article aims to present a real-time monitoring system to detect fires in establishments and on city streets. This system makes use of Machine Learning techniques, more specifically YOLOv5, for fire detection in images coming from IP
cameras via RTSP protocol. For the training, a database was created with annotations specifically for this project. Training was performed on a local machine and the result was a model with a mAP50 of 0.577 for the validation set and 0.413 for the test set. In the test set, a precision of 0.514 and a sensitivity of 0.448 were obtained, which means that it has many occurrences of false positives and false negatives in the detections. Finally, the developed management system – server, web application and database structure – has features such as user registration, authentication, IP camera management, camera image viewing and real-time response to fire detection.
Descrição
Palavras-chave
Machine learning, Sistema, Gerenciamento, Incêndio, Combate a incêndios, Tempo real, Software, Management, Fire, Firefighting, Real time
Citação
OLIVEIRA, João Vitor Almeida de. Desenvolvimento de um software identificador de incêndios utilizando um modelo de machine learning. 2023. 19 f. Trabalho de Conclusão de Curso (Bacharelado em Engenharia de Computação) - Escola de Engenharia Elétrica, Mecânica e de Computação, Universidade Federal de Goiás, Goiânia, 2023.