A New Efficient-Attention Based Disaster Classification for Emergency Monitoring

Natural disasters such as earthquakes, floods and storms have increased today due to climate change. Since infrastructure problems often occur in places where natural disasters occur, it is vital to reach the disaster area as quickly as possible. In such disaster situations, unmanned aerial vehicles can be used, which have significant advantages with their flexibility, low cost and mobility features. In this study, a low-weight deep learning approach that can work on UAV is presented to detect the disaster situation and perform emergency response. The proposed approach is based on the EfficentNetv2 model, which has offered significant advantages in recent years, especially in terms of speed and accuracy. By adding an attention mechanism to this model, its performance has been increased. Experimental results showed that the proposed approach achieved around 6% accuracy compared to the base EfficentNetv2S model.