Select to view content in your preferred language

How to re-use Deep Learning polygons for a different imagery layer

312
3
Jump to solution
01-06-2025 05:24 AM
Labels (2)
ScarletMaguire
Occasional Contributor

I have several different raster layers; each layer is derived from the same scene and each represent a different Vegetation Index layer (where each pixel represents a unique vegetation index value). I wish to use each imagery layer to create Deep Learning samples using the 'Label Objects for Deep Learning' tool. Critically, I wish to re-use the polygon shapes created for one raster layer for each of the raster layers. In other words, I need to create a set of Deep Learning samples using one raster layer, and then simply take the same polygon shapes for those training samples and overlay them onto the other raster layers to create a new set of training samples for those layers. I have experimented with the tools and have not found a way to do this. Can anyone recommend a workflow or suggest a workaround?

 

0 Kudos
1 Solution

Accepted Solutions
David_McRitchie
Esri Regular Contributor

Hey Scarlet

Ah I see, I believe you are using the Deep Learning tools embedded on the Imagery toolbar?

If you go to Analysis  > Tools and use the Export Training Data for Deep Learning tool you can run this as a batch tool to cycle through different Raster Images while keeping all other parameters the same, including the polygons used.

 

Hope that helps,

David

 

Esri UK -Technical Support Analyst

View solution in original post

3 Replies
David_McRitchie
Esri Regular Contributor

Hey Scarlet,

If we are looking to replicate the same workflow, just changing the input to be a different set of polygons then it might be best to try running this within a model or a Python Notebook. These should let you make a variable for the polygons and to reuse the same data.

 

The following documentation might be helpful, please do reach out if you get stuck at any point.

 

Hope that helps,

David

Esri UK -Technical Support Analyst
0 Kudos
ScarletMaguire
Occasional Contributor

Hi David

Thanks for your suggestion. However, what I am looking for is to create several different sets of deep learning training samples using the same set of polygons with the only difference being the input raster/imagery layer. As the 'Label Objects for Deep Learning' tool requires the imagery layer to be first selected, I have not yet found a way to do this. Further, it seems that when I use the tool, I cannot load in deep learning training samples (or the associated polygons layer) created for another imagery layer. But thanks for your suggestion, I will read through the documentation you suggested.

0 Kudos
David_McRitchie
Esri Regular Contributor

Hey Scarlet

Ah I see, I believe you are using the Deep Learning tools embedded on the Imagery toolbar?

If you go to Analysis  > Tools and use the Export Training Data for Deep Learning tool you can run this as a batch tool to cycle through different Raster Images while keeping all other parameters the same, including the polygons used.

 

Hope that helps,

David

 

Esri UK -Technical Support Analyst