---
license: mit
---


# GeoJSON Dataset for Building Damage Assessment

This repository contains the GeoJSON dataset used to fine-tune the UNet model with resnext50_32x4d backbone for building damage assessment on LA fire of January 2025. The dataset is referred to the Eaton fire area and provides detailed annotations of buildings and their damage status.


<div align="center">
  <img src="dataset.png" alt="geojson_dataset_LA_fire_jan_2025" width="800">
</div>


## Dataset Overview

- **Format**: GeoJSON
- **Classes**:
  - Background
  - Building (Healthy Building)
  - Damaged Building

## How to Use the Dataset (however you want)

You can use the dataset coupled with Maxar products available [here](https://xpress.maxar.com/?lat=0&lon=0&zoom=3.0) to finetune other models. You can also expand on this dataset in case the annotations are not enough for your use case.