Deep Learning for Archaeological Object Detection on LiDAR: New Evaluation Measures and Insights

Machine Learning-based workflows are being progressively used for the automatic detection of archaeological objects (intended as below-surface sites) in remote sensing data. Despite promising results phase, there is still a lack standard set measures to evaluate performance object methods, since buried sites often have distinctive shapes that them aside from other types included mainstream datasets (e.g., Dataset Object deTection Aerial images, DOTA). Additionally, research relies heavily on geospatial information when validating output an procedure, type not normally considered regular machine learning validation pipelines. This paper tackles these shortcomings by introducing two novel evaluation measures, namely ‘centroid-based’ and ‘pixel-based’, designed encode salient aspects archaeologists’ thinking process. To test their usability, experiment with different deep neural networks was conducted LiDAR dataset. The experimental show closely resemble semi-automatic one currently archaeologists therefore can be adopted fully detection. Adoption will facilitate cross-study comparisons close collaboration between researchers, which turn encourage development human-centred tools.