A Lightweight Transfer Learning-Based Model for Building Classification in Aerial Imagery

Over the past decade, there has been a growing interest in the potential of artificial intelligence and computerr vision in tackling challenges related to disaster resilience in urban communities. Unmanned aerial imagery has been the focal of a number of initiatives targeting urban planning and aftermath disaster assessment. Within this context, this presents a novel lightweight transfer learning-based model for assessment of building conditions from aerial images. The proposed method is suitable for EDGE-based operations in resource-limited settings. Experiments were conducted to identify post-flooding building conditions in Zanzibar city in Tanzania. Considerable gains in terms of memory and computation time have been achieved while maintaining accuracies that are in line with state-of-art approaches.