AI Suite: Difference between revisions

From Tygron Preview Support Wiki
Jump to navigation Jump to search
Maxim@tygron.com (talk | contribs)
Created page with "The Tygron AI Suite consist of a several tools that can help when using an existing Neural Network or creating a new ope."
 
Maxim@tygron.com (talk | contribs)
No edit summary
Line 1: Line 1:
The Tygron AI Suite consist of a several tools that can help when using an existing [[Neural Network]] or creating a new ope.
The Tygron AI Suite consist of a several tools that can help when using an existing [[Neural Network]] in an [[Inference Overlay]] or when creating a new one.
 
==Steps for creating you own model==
In order to create your own AI model based on a [[Neural Network]] you have to follow these steps.
 
* Start with a definition of which objects you want to detect. For example trees, cars, solar panels, etc.
* Now create one or more projects (with a good variation) and manually create your training data by creating areas containing outlining these objects. See this how to how this can be done in QGIS.
* Devide these objects in two group a TRAIN and TEST dataset that can be exported. See thi manual.
* After exporting follow these steps to train your [[Neural Network]] resulting in a [[ONNX]] file.
* Then import the [[ONNX]] back into the [[software]] an run it in an Inference Overlay.
* Finally validate the results on a different project and iterate back to a previous step if needed.

Revision as of 08:38, 15 October 2025

The Tygron AI Suite consist of a several tools that can help when using an existing Neural Network in an Inference Overlay or when creating a new one.

Steps for creating you own model

In order to create your own AI model based on a Neural Network you have to follow these steps.

  • Start with a definition of which objects you want to detect. For example trees, cars, solar panels, etc.
  • Now create one or more projects (with a good variation) and manually create your training data by creating areas containing outlining these objects. See this how to how this can be done in QGIS.
  • Devide these objects in two group a TRAIN and TEST dataset that can be exported. See thi manual.
  • After exporting follow these steps to train your Neural Network resulting in a ONNX file.
  • Then import the ONNX back into the software an run it in an Inference Overlay.
  • Finally validate the results on a different project and iterate back to a previous step if needed.