Mustatil 5.6 remains free forever
The complete Mustatil 5.6 release stays available as a permanent free edition.
AI-powered geospatial analysis
Mustatil 6
A complete desktop workspace for large geospatial imagery. Detect, annotate, train, segment, review, and export results without splitting your workflow across separate applications.
GeoTIFF & BigTIFF
YOLO & Faster R-CNN
SAM2 segmentation
AI pipelines
Mustatil 6 is available for purchase
Get the newest generation through the Microsoft Store and support continued development.
One workspace from detection to review
Mustatil 6 combines large-raster inference, satellite imagery workflows, video analysis, annotation, training, segmentation, and GIS-ready export.
Mustatil is built for projects where conventional computer-vision tools become impractical: huge GeoTIFF and BigTIFF files, satellite mosaics, remote-sensing imagery, archaeology, geospatial surveys, and long-running tiled detection jobs. The workspace keeps model selection, inference, annotation, quality review, and export together in one desktop interface.
Mustatil means rectangle — a reference to both archaeological mustatils and the rectangular detection boxes used in AI object detection.
Mustatil 6 features
A broader workspace for geospatial AI, image intelligence, dataset preparation, model development, and operational review.
- Object detection in huge GeoTIFF and BigTIFF datasets
- Memory-conscious tiled processing with overlap control
- Satellite imagery detection and remote-sensing workflows
- YOLO, Faster R-CNN, RF-DETR, Mask R-CNN, U-Net and DINO support
- Video detection, annotation, timeline review, and export
- SAM2-assisted segmentation from boxes and annotations
- Integrated annotation, class management, and dataset creation
- Graphical AI pipeline workflows and model chaining
- Detection filtering, scoring, and FormTrainer integration
- Exports to GeoJSON, GeoPackage, CSV, images, masks, and datasets
AI model support
Use fast detectors, proposal-based networks, transformer architectures, semantic segmentation, interactive masks, and prompt-driven workflows.
YOLO / Ultralytics
Tiled detection, training, class filtering, review, model comparison, and GIS export for Ultralytics-compatible models.
Faster R-CNN
Proposal-based detection with selectable backbones and large-image tiled inference.
RF-DETR
Modern transformer-based object detection as an alternative to conventional one-stage detectors.
Mask R-CNN
Instance segmentation for projects that require object masks in addition to bounding boxes.
U-Net
Pixel-level semantic segmentation for remote sensing, raster classification, and image-to-mask workflows.
SAM2
AI-assisted segmentation using annotation boxes as prompts, with single-image and batch processing.
Grounding DINO
Text-guided detection and flexible prompt-based localization beyond fixed training classes.
OWL-ViT / OWLv2
Experimental open-vocabulary detection for locating concepts from text prompts.
LAE-DINO and advanced workflows
Project-based training, detection, experimentation, filtering, and model-assisted review.
Downloads and editions
Purchase Mustatil 6 from Microsoft or keep using the permanent free Mustatil 5.6 release.
Mustatil 6 — Microsoft Store
The current generation of Mustatil. Available for purchase from the Microsoft Store.
Mustatil 5.6 — Free forever
The complete Windows desktop installer remains permanently free.
Mustatil 5.6 for macOS
Free Apple macOS package from the Mustatil 5.6 release.
Mustatil 5.6 for Linux
Free Debian package from the Mustatil 5.6 release.
Python package
Copy and paste this command into a terminal to install Mustatil from PyPI.
pip install mustatil
Project links
Source code, releases, documentation, and alternative downloads.
Mustatil 6 screenshots
Real workspace views showing large-raster detection, Faster R-CNN inference, video intelligence, and SAM2 segmentation.
Standalone AI trainers
Dedicated Microsoft Store applications are also available for focused training workflows outside the complete Mustatil workspace.
YOLO Trainer
Dedicated training application for creating and managing YOLO object-detection models.
R-CNN Trainer
Standalone application for region-based object-detection training and experimentation.
RF-DETR Trainer
Focused RF-DETR model-training workflow for transformer-based object detection.
More applications by Tarek Wasfy
Browse the complete Microsoft Store publisher catalog.
Scientific archive
Permanent Zenodo reference for citation and archival access.
Additional information
GIS compatibility and project notes.
GeoPackage results can be opened in QGIS and other GIS applications. Layer CRS and EPSG settings can be adjusted directly in the GIS software when required.
The program was written using AI.