Under development

SeismicBuildingExposure

SeismicBuildingExposure is a Python package that computes everything you need to estimate the seismic exposure of buildings using only geospatial data and no field surveys.

---

Installation

To install the package, use the following command:

pip install git+https://github.com/GeomaticsCaminosUPM/SeismicBuildingExposure.git

---

Modules

1. MLfootprint

Machine learning AI model for automatic instance segmentation of building footprints.

• Fine tune the SAM2 model for assisted segmentation.
• Fine tune the maskformer model for automatic segmentation.
• Run new images (interference) on the SAM2 model for assisted segmentation.
• Run new images (interference) on the maskformer model for automatic segmentation.

To create and download a dataset from publically available sources use the data module.

2. MLstructural_system

Predicts the structural system of a building using the data extracted from the footprint, height and remote_sensing modules.

• Fine tune or train your own bayesian model with your own survey.
• Predict using one of our pre-trained model on your own dataset.

3. data

The data module is based on the GeoVisionDataset library and provides functions adapted to create building footprint datasets using publically available datasources.

4. fooptrint

This module is divided in 2:

• position: Relative position of the building (in a row, on a corner, isolated, etc.)
• shape: Footprint shape and irregularity according to international building codes.

5. height

Provides functions to process a .ply point cloud and get building heights and other height related data.

• DSM: Digital surface model (raster image) form a .ply point cloud.
• DTM: Digital terrain model (raster image) using the DSM as input and considering ground points to be on streets downloaded from OpenStreetMap.
• height: Building height, roof steepness, ground altitude and ground steepness on every building footprint.
• irregularity: Irregularity in elevation according to international building codes.

6. remote_sensing

• photogrammetry: COLMAP open source photogrammetry workflow to get a .ply dense point cloud from drone images.
• roof_material: Estimation of the roof material using LANDSAT and SENTINEL-2 spectral information.
• year: Estimation of the year of first construction and year of last building change using LANDSAT sattelite imagery.