QGIS GeoAI 2.0

The new version of the QGIS GeoAI plugin is ready and available for use by the GIS user community worldwide.

This new version constitutes a significant improvement in functionality, however, it does not require major configurations, if you have installed and used the previous version of GeoAI simply replace the plugin.

Following the guidelines expressed in the article How to obtain better results when capturing elements on geospatial images? On this occasion I have focused on: exploiting the unknown or unimplemented capabilities of the SAM model.

What changes does this new version of GeoAI introduce?

The following is a description of the improvements introduced in each GeoAI module in version 2.0.

Interactive Segmentation

1.- The options for the extraction of elements with multiple capture are extended.

• By selecting the Multiple Capture option you can extract an element using multiple points to select it.

• Using the drop-down list Prompt Type Positive/Negative you can use points with positive (include) and negative (exclude) selection.

• You can combine one selection per area with multiple selection points.
• The original version of SAM and SAM Ligth support all these features.

2.- Other improvements

• Clicking on the Advanced Options button provides a button to create a vector layer with the selection points and areas.

Segmentation of the whole image

• The Pre-loading process of the model/image separates, now it must be executed in the dialog box of the Segment whole image module.

• You can now use the Light version of SAM to segment the entire image. This is a great new feature, allowing you to segment an image with lower power hardware or over larger images, as well as combine with the advanced configuration options.
• By clicking on the Advanced Options button, the configuration alternatives are displayed. Options have been added to allow the use of GPUs with low memory, even less than 4 Gb.

• The option Use point layer (EXPERIMENTAL) is a feature under development, although it should not generate conflicts or errors, it will not be useful for the moment.

How to access and install the GeoAI plugin?

You can download it and consult the documentation with installation details in the Github repository.

Endnotes

Going into more details about the functionality of GeoAI version 2.0 is beyond the scope of this article.

In future publications and by updating the documentation in the Github repository, I will go deeper into usage and performance aspects.

Especially in the course that I will be organizing soon we will go into more details on how to take full advantage of this AI model in the geospatial field.

The development of this new version was done on QGIS version 3.28, if you have errors when testing it on other versions of QGIS, I appreciate you communicate it to make the necessary adjustments.

What's next?

The objective of the article How to obtain better results when capturing elements on geospatial images? was to draw a roadmap towards which to guide the development, as you will notice there are many alternatives for improvement, highlighting the incorporation of tools to adapt the images to the model before processing.

Why continue to improve GeoAI?

• The use of AI in the geospatial field is a developing topic, it has great potential, but it requires a sequential and comprehensive approach, without neglecting the objectives and the integration with classic GIS and Machine learning tools.
• Without such tools, AI models in GIS would be available only in commercial applications and with very powerful computing equipment.
• On the other hand, although the functions incorporated in GeoAI significantly improve the capture process and allow its implementation by any user, there are still images and conditions in which the result can be improved.

Support the development

If you consider that this tool is useful and you want to promote its development, any contribution or collaboration is welcome, you can write to me privately and we will coordinate it.