ERROR 001109 in deep learning

[SolanaFoo2]

Getting the same error on a model I just created in Pro.  What is this "inPropertyValue; buffer ..."?  Also getting this error from the dlpk that came with this ESRI tutorial: https://learn.arcgis.com/en/projects/automate-fire-damage-assessment-with-deep-learning/

ERROR 001109 Unrecognized token type; state : inPropertyValue; buffer : : Infinity,
"ModelFile": "buildings_classify.

[PavanYadav]

Hi  

I tried to reproduce the error using  tutorial: https://learn.arcgis.com/en/projects/automate-fire-damage-assessment-with-deep-learning/ but the tool worked for me. We would like to investigate this further and wondering if you can please share the following info? 

1. What version of ArcGIS Pro are you using? 

2. How did you install the Deep Learning Packages e.g. from installer or the command as documented here: https://github.com/Esri/deep-learning-frameworks

3. Does this error occurs with other models or just this one? 

4. Can you try the emd file  instead of the dlpk as a model in the tool and see if it works? 

Thanks

Pavan

[PavanYadav]

Can you also share what was your Windows/ArcGIS Pro languages and local set to? 

[AyeshaM]

Hi, I am also facing this issue 

I am trying classify pixels using deep learning in Arc Pro 3.0 but the tool is not accepting the .emd file or .dlpk file and throws this error:

Unrecognized token type; state : inArrayValue; buffer : , Infinity ], "band

Thank you in Advance

[PavanYadav]

@AyeshaM Can you share the .emd and \ModelCharacteristics\training_validation_loss.json? 

thanks 

Pavan

[Hanine_Serhan]

Hi,

How many bands your image has? can you please share your Raster info and band metadata?

I faced the same issue when working with 4 bands imagery with bad values in the 4th one

I used to extract RGB bands using  the "Extract Bands" Raster Function, and use the output raster to train my model

[Wei-HsinFu]

I encountered the same problem in Pro 3.1. The problem went away after I replaced the 32-bit raster used in the training data by the 8-bit signed version. Learned from a video that the 8-bit rasters work best with the ArcGIS deep learning tools.