Issues with Classify Pixels Using Deep Learning in Experience Builder Analysis Widget (Enterprise 11.4)

[Muhammad_Elbagoury]

Hi everyone,

I’m working in ArcGIS Experience Builder and trying to use the Classify Pixels Using Deep Learning tool through the Analysis Widget.

Here’s what I’ve done so far:

• I downloaded the High Resolution Land Cover Classification - USA deep learning package (DLPK) from the Living Atlas:
  https://www.arcgis.com/home/item.html?id=a10f46a8071a4318bcc085dae26d7ee4

• I uploaded the deep learning model to my ArcGIS Enterprise Portal.

• I set up an Experience using the Analysis Widget configured to run Classify Pixels Using Deep Learning, pointing to the uploaded model.

However, I am encountering two issues:

1. When loading the model from My Content, I get this error:
   "Failed to query model arguments." (Issue_1)

2. When I try to run the tool, I encounter:
   "Error: Cause of failure: Failed to execute. Parameters are not valid." (Issue_2)

Some additional information:

• ArcGIS Enterprise version: 11.4

• Raster input: Stored in ArcGIS Online as an image tile service.

Has anyone faced this issue before or have any suggestions to fix it?
Thanks a lot in advance!

[Faiez]

Hi!

I have tested with ArcGIS pro. The problem occure if the version of deep learning framworks is not the same as your ArcGIS Pro. If the version of your ArcGIS pro and framwroks are the same, It works smoothly. Check your ArcGIS Experience builder version and your Deep learning framworks. It does not works, if you do not install the Deep learning framworks. You have to download the version of Deep Learning framwroks from github from the below link.

https://github.com/Esri/deep-learning-frameworks

[Muhammad_Elbagoury]

Hi @Faiez ,

Thank you so much for your response and for sharing the details! I really appreciate your help.
Just to clarify: I already have the Deep Learning Libraries Installer for ArcGIS Server 11.4 installed, as I’m working with ArcGIS Enterprise 11.4. Also, I can confirm that the workflow runs smoothly for me when using ArcGIS Pro 3.4 together with the Deep Learning Libraries Installer for ArcGIS Pro 3.4.

My issue is specifically with ArcGIS Enterprise.

Thanks again for your support!

[Faiez]

Salam Muhammad!

You said, “when you try to run the tool, you encounter
Error: Cause of failure: Failed to execute. Parameters are not valid. (Issue_2)”

It clarify that you have not reprocessed the pixel values, this mean that the training and the testing imagery does not share the same parameters, such as pixel size, pixel depth and band and maybe coordinates. It is important to make sure that samples in the same class have the same class, code, name and value, class value cannot be 0. The DL classifying pixel model must be unzipped. From DLPK you have to choose only the model that required carrying the operation.

What model have you tried for classifying pixels that show the error?

Running deep learning in an ArcGIS online or Ex builder application consumes credits, while ArcGIS Pro friendly environment does not consume credits.

Sometimes, it is good to uninstall the Deep Learning Library and reinstall it, and make sure it has been removed from the folder.

Good luck.

[Wei_Ying]

@Muhammad_Elbagoury thanks for you feedbacks!
Can you try from the same tool in map viewer - Analysis and see if you have the same issue?

[Muhammad_Elbagoury]

@Wei_Ying Thank you for your response.

I have tested two tools within the ArcGIS Online (AGOL) environment.

• The "Detect Objects Using Deep Learning" tool ran successfully in both Map Viewer and Experience Builder.
• The "Classify Pixels Using Deep Learning" tool, however, failed in all four attempts when tested with two different models across Map Viewer and Experience Builder.

Based on these results, I believe there may be an issue specific to the Classify Pixels Using Deep Learning tool.

Thank you again for your support.

[Wei_Ying]

@JieZhang could you please help take a look? Seems an issue with the Classify Pixels Using Deep Learning tool.

[Muhammad_Elbagoury]

Hi @Wei_Ying,

I also tested the same workflow in my ArcGIS Enterprise environment, and the results were different from AGOL.

• The "Detect Objects Using Deep Learning" tool ran successfully in both Map Viewer and Experience Builder as AGOL.
• The "Classify Pixels Using Deep Learning" tool failed in both applications when tested with two different models. However, unlike in AGOL, the tool in ArcGIS Enterprise did not load the model parameters at all, which prevented the tool from running.

I’ve attached screenshots to help illustrate these issues. Thank you again for your continued support.