• Informações do Trabalho

Titulo Fire detection through a combination of deep neural networks and graph cuts Subtítulo Autor DAVI MAGALHÃES PEREIRA Orientador SAULO MORAES VILLELA Resumo Recent developments in computer vision techniques have markedly improved fire detection capabilities compared to conventional systems. This work introduces an innovative methodology that integrates deep neural networks for identifying instances and regions of fire, graph cuts, and color thresholding for a nuanced approach to fire segmentation. The incorporation of fire segmentation masks facilitates precise analysis, providing valuable insights into fire origins and propagation to proactively prevent future incidents. Our method, leveraging graph cuts segmentation with comprehensive color information, demonstrates enhanced accuracy and detailed fire detection. The results illustrate a notable improvement in recall, maintaining competitive precision, thereby establishing an efficient and effective fire detection framework. Palavras-Chave Fire detection; fire segmentation; image classification; graph cut; deep learning; color thresholding.