samples:samples:aimodelrunner
Table of Contents
AIModelRunner sample
Documentation | Samples | AIModelRunner
Summary
Sample app for running Python-driven AI Models on TatukGIS Viewer/Layer.
Description
What the sample shows:
- Loading image files (TIFF, etc.) into the GIS viewer
- Executing AI/ML models on raster data via Python backend
- RealEsrgan model for image upscaling (super-resolution)
- MMRotate model for object detection and orientation
- Selecting models from dropdown list
- Background model execution with progress feedback
- TGIS_AIPythonWorker for asynchronous model processing
- TGIS_AIModelCustom for wrapping custom AI implementations
- Dual-target .NET build (net6.0-windows + net4.8)
- GisImageExporter for output raster generation
- Handling long-running AI computations without UI blocking
- Integration with Python environment and model libraries
What the sample shows:
- Loading image files (TIFF, etc.) into the GIS viewer
- Executing AI/ML models on raster data via Python backend
- RealEsrgan model for image upscaling (super-resolution)
- MMRotate model for object detection and orientation
- Selecting models from dropdown list
- Background model execution with progress feedback
- TGIS_AIPythonWorker for asynchronous model processing
- TGIS_AIModelCustom for wrapping custom AI implementations
- Dual-target .NET build (net6.0-windows + net4.8)
- GisImageExporter for output raster generation
- Handling long-running AI computations without UI blocking
- Integration with Python environment and model libraries
This sample demonstrates how to execute AI inference models (like Real-ESRGAN or custom ONNX/PyTorch models via Python wrappers) directly from TatukGIS. It exports the map state, executes a Python script, parses a JSON manifest, and maps the output (raster layers or vector detections) back onto the map.
Check following guide: Running AI Models in Developer Kernel to learn how to run custom AI models in TatukGIS DK environment.
This sample illustrates use of:
| Name | Description |
|---|---|
| PythonWorkerDefault | Get or create the default singleton worker and start it immediately. |
| TGIS_AIPythonWorker.Enqueue | Queue a job for execution on the worker thread and return immediately. |
| TGIS_AIModelCustom.Create | Instantiates a custom AI model with a Python manager, model path, script template, required modules, and placeholder tokens. |
| TGIS_AIModel.InstallModules | Installs the Python modules required by the model. |
| TGIS_AIModel.Run | Runs the model for the specified input image. |
| TGIS_AIModelOutput.GetDetections | Returns detections aggregated from all detection output items. |
| TGIS_AIModelOutput.GetDetectionLayers | Returns detections converted into vector layers. |
| TGIS_AIModelOutput.GetLayers | Returns file-backed layers produced by the AI output manifest. |
| TGIS_ImageExporter.ExportToImage | Export viewer content to an image file using TGIS_PixelExportManager. |
| TGIS_ImageExporter.GetPixelSize | Calculate how a single pixel translates to geographic distance from a source viewer, based on its visible extent and viewer component size. |
Code
This sample is available on the following platforms (click to view the source code on GitHub):
2026/07/05 05:38