Reproduce “Archaeoscape: Bringing Aerial Laser Scanning Archaeology to the Deep Learning Era” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yohann Perron, Vladyslav Sydorov, Adam P. Wijker, Damian Evans, Christophe Pottier, Loic Landrieu
Researchers created Archaeoscape, a large-scale, open-access dataset of airborne laser scanning (ALS) data with expert annotations of archaeological features from the Angkorian period in Cambodia. This resource facilitates the application and benchmarking of deep learning models for automated archaeological discovery in densely vegetated environments.
Key insights
- Elevation data (nDTM) is critically important for archaeological feature detection from ALS, especially in vegetated areas, significantly outperforming radiometric (RGB) data alone.
- Convolutional Neural Networks (CNNs) and Hierarchical Vision Transformers (HViTs) that integrate local feature processing demonstrate competitive or superior performance on ALS archaeology tasks compared to standard Vision Transformers.
- Incorporating a larger spatial context through increased input image sizes (e.g., 512x512 pixels) improves model performance for detecting large-scale archaeological features.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.