DeepDamageNet: A two-step deep-learning model for multi-disaster building damage segmentation and classification using satellite imagery

TL;DR

This work tackles rapid post-disaster building damage assessment by jointly analyzing pre- and post-disaster satellite imagery. It compares an end-to-end approach with two-step pipelines, finding that separating building segmentation from per-building damage classification yields better performance, especially when incorporating disaster-type priors. The best two-step model with disaster-label information achieves an overall F1 score of about 0.66, substantially exceeding the baseline of 0.28, underscoring the benefits of task decomposition and prior information. The results highlight ongoing challenges in cross-disaster damage classification and suggest probabilistic priors and disaster-aware strategies as key directions for practical deployment.

Abstract

Satellite imagery has played an increasingly important role in post-disaster building damage assessment. Unfortunately, current methods still rely on manual visual interpretation, which is often time-consuming and can cause very low accuracy. To address the limitations of manual interpretation, there has been a significant increase in efforts to automate the process. We present a solution that performs the two most important tasks in building damage assessment, segmentation and classification, through deep-learning models. We show our results submitted as part of the xView2 Challenge, a competition to design better models for identifying buildings and their damage level after exposure to multiple kinds of natural disasters. Our best model couples a building identification semantic segmentation convolutional neural network (CNN) to a building damage classification CNN, with a combined F1 score of 0.66, surpassing the xView2 challenge baseline F1 score of 0.28. We find that though our model was able to identify buildings with relatively high accuracy, building damage classification across various disaster types is a difficult task due to the visual similarity between different damage levels and different damage distribution between disaster types, highlighting the fact that it may be important to have a probabilistic prior estimate regarding disaster damage in order to obtain accurate predictions.

Figures (10)

• Figure 1: Pre-disaster (left) and its corresponding post-disaster satellite image
• Figure 2: Distribution of building damage level in training set
• Figure 3: Differences in building densities
• Figure 4: Architecture of two-step model. The pre-disaster image is fed into a segmentation model to obtain building polygon coordinates. This result in addition to pre- and post- disaster cropped building image is then fed into a damage classification model to obtain the final image.
• Figure 5: Architecture of end-to-end model. The pre- and disaster image is subtracted before feeding into a U-Net to perform simultaneous segmentation and classification.