Mustatil Yolo AI Trainer

Mustatil YOLO AI Trainer is an open-source AI object-detection tool for training and running detectors on satellite imagery, maps, and very large images at GIS-level precision. It is aimed at archaeologists, remote-sensing researchers, and GIS professionals who need custom detection on geospatial datasets rather than everyday photos. The workspace pairs a class-based annotation workflow for building training sets with YOLO/Ultralytics-based training and inference, and it can export detections to GeoPackage and other GIS-compatible formats for downstream mapping. Beyond YOLO, it supports multiple model families including DETR, Faster R-CNN, Mask R-CNN, U-Net, and SAM2, and offers a visual pipeline builder plus map-based review so you can correct detections on a geospatial canvas. The desktop app is built in Python with a PyQt interface and handles GeoTIFF and BigTIFF formats. It is distributed under LGPL v3 as free, cross-platform software (version 5.6), with installers for Windows, macOS, and Linux, a PyPI package, and a Snap Store option; a Windows-only C++ GUI is planned via the Microsoft Store for $4.99. As a research-grade geospatial detection trainer, it expects familiarity with GIS concepts and model training.

Training and running object-detection models on satellite imagery and maps, with GIS-format export for archaeology and remote sensing