SVM Train Classifier tool failing

[MichaelHuffman]

I am continuing to run into an issue when trying to train the classifier in the Image Classification Wizard workflow. The wizard completes the segmentation but as soon as it starts training the classifier the tool fails. I cannot figure out what is causing the failure as the only error being returned is the 999999. Any guidance help is greatly appreciated.

Here is a copy of the report:

    Training Classifier...
    Start Time: Wednesday, August 12, 2020 3:45:41 PM
    ERROR 999999: Something unexpected caused the tool to fail. Contact Esri Technical Support (http://esriurl.com/support) to Report a Bug, and refer to the error help for potential solutions or workarounds.
    Failed to execute (TrainSupportVectorMachineClassifier).
    Failed at Wednesday, August 12, 2020 3:45:41 PM (Elapsed Time: 0.40 seconds)

Thanks!

Mike

[ChrisBiondi]

I am having the exact same issue. If you ever figured out any work arounds let me know. Seems like just another bug. 

Thanks!

[AndrewBartle]

Same issue here. Please let me know if you figured anything out! Thanks

[Pål_Herman_Sund]

I have the same issue with a client. The strange thing is I am not able to reproduce, neither via the wizard or by doing the steps with the GP Tools and the Training Manager. And he has a computer muvh more powerful than mine.. This case is escalated to Esri and we will have to see the outcome. Image is a pretty standard 3-band 8-bit RGB ortophoto. 

[BrettLawrence]

Did you ever get this resolved? I am having the same issue.

[Pål_Herman_Sund]

'My' escalated case has been up to top-level but dev not able to reproduce either. At my client side it is a pretty 'strict' IT-environment and a wild guess would be the environment plays a role. 

[Pål_Herman_Sund]

@MichaelHuffman @BrettLawrence @ChrisBiondi @AndrewBartle we sort of reached a conclusion in my case. I am working at the Norwegian distributor supportservices and was never able to reproduce this problem on the same data as my client used (standard 3 band, 8-bit RGB imagery - but a pretty big one..). But our client had all kind of problems.. Just now their internal "raster expert" arrived from 1 year leave and he was able to classify the very same imagery on the very same workhorse of a computer at the client site. But he had some more privileges on his AD user, and that made all the difference. Exactely which privileges I never got to track down and it is a little "out of scope". 

There is one similar case which has arrived at my desk, not finished troubleshooting, but I think the python environment is the problem in that case