Vortrag auf Einladung der ÖGUF:
The talk brings together airborne laser scanning (ALS, lidar), relief visualisation, and deep learning to identify and classify archaeological features in landscapes. Visualisation products derived from raster elevation models remain the foundation of most archaeological analyses of ALS data. The Relief Visualisation Toolbox (RVT) was developed to support the visual interpretation of elevation-model datasets through a curated set of methods that have proven effective for detecting small-scale features, with default parameters tuned for that purpose. RVT is now available as a Python library, a QGIS plugin, and as raster functions for ArcGIS Pro and ArcGIS Enterprise, allowing these techniques to be easily computed for both manual and automated interpretation.
An archaeological workflow for building machine-learning models will be presented, alongside experiments covering a range of contexts: buildings, platforms, and aguadas of the ancient Maya; enclosures, ringforts, and barrows in Ireland; and a landscape of barrows in south-eastern Herzegovina. The underlying models are trained on an extensive archive of ALS datasets labelled by domain experts. A series of experiments compares visualisation methods and machine-learning architectures for object detection and semantic segmentation, with the goal of identifying optimal configurations for a practical, user-friendly tool for Automatic Detection of Archaeological Features (ADAF).
Zugangsdaten:
Teilnahme über Zoom
https://us06web.zoom.us/j/85934753909?pwd=e2PEboknlVZvFxgiloXsOGTXwa8U1d.1
Teilnahme über den Internet-Browser:
https://zoom.us/join
Meeting-ID: 859 3475 3909
Kenncode: 590203