An improved stacked sparse autoencoder technique for victim detection using deep learning based multimodal imagery

Puntos clave

• The approach achieves higher accuracy in victim detection across various scenarios, improving existing techniques.
• Key metrics indicate increased detection rates of up to 30% compared to traditional methods within testing datasets.
• The method utilizes a stacked sparse autoencoder framework, integrating data from different image sources for robust analysis.
• This technique highlights the potential for real-world applications in disaster situations, enhancing response efforts.