Supported projects

Recent discoveries have revealed increasing evidence of complex ancient human activities within tropical forests. Remote sensing technologies, such as LiDAR, have uncovered vast and intricate archaeological landscapes beneath the forest canopy. Yet, LiDAR post-processing remains challenging, requiring manual intervention and parameter tuning at critical stages, as existing methods have not been specifically adapted for detecting archaeological features in dense forests. This increases the risk of missing subtle archaeological signatures and limits the effectiveness of deep learning models in automated mapping. We address two fundamental challenges in LiDAR post-processing. First, we develop a specialised machine learning model for automated 3D point cloud segmentation, enabling consistent separation of ground, vegetation, and anthropogenic features. Second, we use this segmentation model to standardise existing datasets, facilitating the development of machine-learning-driven archaeological mapping that can scale across large datasets. This approach lays the groundwork for a foundation model for LiDAR archaeology, while also opening avenues for methods that extend beyond direct LiDAR coverage. The ALiS project builds on EFEO and CASE’s experience in large-scale airborne LiDAR applications, drawing on EFEO’s extensive dataset and ENPC’s strengths in 3D point cloud processing to develop more efficient and scalable methodologies for archaeological analysis. This effort expands on ongoing work in EFEO’s machine learning lab, leveraging its computational infrastructure and an interdisciplinary network of archaeologists, computer scientists, and heritage specialists.