Semantic segmentation dataset on earthquake damage

홈페이지 · 논문 1편

This dataset contains 547 social media images taken in the aftermath of various earthquakes. Each image is paired with a pixel-wise semantic segmentation mask that categorizes the scene into three classes: - Undamaged structures - Damaged structures - Debris The segmentation masks were manually labeled and reviewed to support training and evaluation of machine learning models for pixel-level damage severity estimation.