Suggesting Exit Path in a Multi-Path in Case of an Emergency

This study proposes an innovative approach to address the challenges of navigating multi-exit scenarios during emergencies by utilizing the VGG19 model and incorporating voice guidance. A crucial component of this research is the extensive CubiCasa5K dataset, comprising 5,000-floor plans with detailed annotations, including exit paths and structural information from diverse indoor environments. These floor plans serve as a valuable resource for generating synthetic images of exit paths within various buildings. The synthetic images undergo pre-processing and are inputted into the VGG19 model for feature extraction and exit path classification. To enhance user experience, a text-to-speech synthesis system is integrated, generating a tailored voice message offering clear and concise instructions based on the VGG19’s classification of the exit path. The solution achieves an impressive accuracy of 94.96%, promising to significantly improve individual safety by providing accurate and timely guidance in complex emergency situations.