Post-Disaster Classification of Building Damage Using Transfer Learning

Building damage assessment after natural disasters is an important task for disaster managers and practitioners. In order to provide detailed levels of post-event building damage, this paper applies deep learning models for building localization and damage classification using transfer learning with an online free xBD dataset. The model is pretrained with ImageNet dataset. SE-ResNeXt-50-32x4d is applied for building localization, and HRNet is applied for damage classification. The building damage is divided into four levels, including no damage, minor damage, major damage, and total damage. The results show that the method can be applied for classifying building damage in an acceptable manner. This can help the government and rescue teams make disaster response quickly and support disaster management.