An Attention-Based System for Damage Assessment Using Satellite Imagery
2021
When a disaster strikes, accurate situational information and a fast, effective response are critical to save lives. High resolution satellite images enable emergency responders to estimate the location, cause, and severity of damage. In this paper, we present a Siam-U-Net-Attn model with an attention mechanism to assess the damage level of buildings given a pair of satellite images showing a scene before and after a disaster. We evaluate the proposed method on xView2, a building damage assessment dataset, and demonstrate that the proposed approach achieves accurate damage scale classification and building segmentation.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
8
References
0
Citations
NaN
KQI