X-AnyLabeling Integration Into Mustatil a Free GIS (Near) Universal AI Trainer

[tarekwasfy01]

I have added a download button and a label synchronisation for X-AnyLabeling to Mustatil.

The software supports YOLO, RF-DETR, Faster R-CNN, custom annotation, model training, false-positive labeling, tiled processing of large GeoTIFFs, and export to GeoPackage / GeoJSON for GIS workflows.

My main use case is archaeological landscape analysis, especially detecting possible mustatil structures, cairn-like features, burial mounds, enclosures, and other geometric features in large remote-sensing datasets.

The goal is not to replace archaeological interpretation, but to create a local tool that helps researchers scan large landscapes, train their own models, review detections, remove false positives, and export results for proper GIS analysis.

GitHub: Github
Zenodo: Zenodo

[hexsyro]

Nice integration. The tiled GeoTIFF processing + false-positive review loop is a solid workflow — that human-in-the-middle step is what makes ML actually useful for archaeology rather than just noise. For training data beyond remote sensing (e.g., social discussion around sites, tourism impact, researcher collaboration networks), we build custom datasets at Social Intel if that's ever relevant to your pipeline.