AI-Assisted Detection of Candidate Cairn, Pendant and Mustatil Features near Majami al-Hadb, Central Saudi Arabia, Using Faster R-CNN and Mustatil Software

This dataset contains 8,761 machine-generated candidate detections of possible mustatil-related archaeological features in central Saudi Arabia, near Majami al-Hadb. The detections cover an area between approximately 43.483947°E and 44.368286°E longitude and 20.662341°N and 21.848753°N latitude WGS84. The dataset was produced through an AI-assisted archaeological remote-sensing workflow using a Faster R-CNN ResNet-50 FPN object detection model. The detection results are provided as geospatial vector data in GeoPackage format, preserving polygon geometries and associated confidence values for each candidate feature. The corresponding trained model weights are also deposited with the dataset as a .pth file to support reproducibility, verification, and further reuse. Detection and GIS export were carried out using Mustatil, a desktop software environment for AI-assisted archaeological and geospatial image analysis. Mustatil was used for model-based detection, spatial visualization, candidate review, and export of the detection results into GIS-compatible formats. The detections should be understood as automated candidate features rather than confirmed archaeological sites. Manual validation, comparison with high-resolution imagery, and further archaeological assessment are required before interpretation or publication as confirmed mustatil structures. This work was created with the help of AI tools.