ArchaeoAstroInsight
Detects archaeological features in satellite and aerial imagery and computes their archaeoastronomical alignment by using AI algorithms.
ArchaeoAstroInsight is a QGIS plugin for the archaeoastronomical analysis of ground-level archaeological features. It detects features in satellite or aerial rasters using a YOLO model, estimates each feature's orientation (azimuth) with a ResNet classifier (or reads it directly from user-drawn lines), derives the local horizon from the HeyWhatsThat online service, user-provided SRTM elevation data and/or LiDAR point clouds, and computes the astronomical declination of each alignment. It also generates azimuth and declination curvigrams (kernel-density plots with significance thresholds) and writes all results to a single CSV for further analysis. It requires external Python packages (PyTorch, TensorFlow, Ultralytics and others) that are not bundled with the plugin; download the installer scripts and follow the installation instructions in the repository README (https://github.com/geo-andrei/ArchaeoAstroInsight). Data to reproduce the experiments is available as a Kaggle notebook: https://www.kaggle.com/code/marcfrincu/al-hayit-analysis