Deep Learning to Assess Palm Tree - prepare_data problem

6610
34
11-12-2019 05:32 AM
MichalRuš
New Contributor II

Hello,

this is my first post on this forum. I'm new to the concept of Deep Learning and I'm following the lesson "Use Deep Learning to Assess Palm Tree Health". I was able to follow this tutorial without a problem until I got stuck on prepare_data function with the following error:

prepare_data function error message

So I followed the suggestion and tried to install those packages again with the following result:

conda installation result

However that did not help. So I tried to install the latest versions of those packages as following:

That did not help either. So now I'm asking what am I doing wrong ?

My PC specs:

CPU: AMD Ryzen Threadripper 1950X

GPU: NVIDIA GTX 1080 Ti with latest drivers

OS: Win 10 64bit

Latest NVIDIA CUDA Toolkit and cuDNN package installed.

I use ArcGIS Pro 2.4.2 trial

I haven't tested the Deep Learning tools in ArcGIS Pro yet. I suppose I have all necessary HW and SW installed to run either Pytorch or TensorFlow on my GPU but I don't want to do this before this problem is solved. Any possible testing would not make sense anyway without solving this issue. Anyone else got this problem ?

Thanks !

deep_learning #palm_trees prepare_data

34 Replies
MichalRuš
New Contributor II

Thanks for the suggestion Shaun ! This really helped !

But while executing the python script in jupyter lab I came to another error:

I believe the error produced by the float conversion function has something to do with a comma (,) symbol in the string value. Is it a problem of arcgis python implementation on Windows ? How can I correct this ?

0 Kudos
ShaunWalbridge
Esri Regular Contributor

Great!

That looks like a bug in the implementation of the Python API. If you're adventurous, you can try fixing with code like this:

python - How do I parse a string to a float or int? - Stack Overflow 

Or (temporarly) change your locale period by using one of the blue countries in this map:

  https://upload.wikimedia.org/wikipedia/commons/a/a8/DecimalSeparator.svg 

I would also report this issue here:

  GitHub - Esri/arcgis-python-api: Documentation and samples for ArcGIS API for Python 

Thanks,
Shaun

0 Kudos
WilliamHarbert
New Contributor III

Hello, I have this exact same error.  I checked the GitHub and the pytorch tutorial is blank without uploaded code.  The tensorflow example from GitHub works, but you must be very careful with the versions installed within python.  Thanks for posting this earlier--it's frustrating eh?

0 Kudos
WilliamHarbert
New Contributor III

I also recreated the python shell from scratch exactly duplicating the tutorial commands.  Same result.

I then completed the conda install as suggested in the error message:

OMG--the above clean install and update fixed this particular error!  YIPPEE!

0 Kudos
WilliamHarbert
New Contributor III

Oops, wrong again. The Kernel was not running.  Note above in the upper right--the No Kernel notification.

I restarted the Kernel and received the same error as earlier. \

0 Kudos
WilliamHarbert
New Contributor III

This has the same error using the Python3 kernel.  Any help would be greatly appreciated!

0 Kudos
WilliamHarbert
New Contributor III

I recalculated everything using the exact capitalization used in the on-line tutorial.  After reading the documentation for prepare_data I added some other attributes--no change.  This is curious.  I also switched to jupyter classic so I could Trust the notebook.

Still the same error.  It seems that the path is never actually used:

0 Kudos
WilliamHarbert
New Contributor III

Slight modification--still an error with prepare_data:

0 Kudos
WilliamHarbert
New Contributor III

This seems like it should work, but fails.  Any ideas would be greatly appreciated!

0 Kudos
VinayViswambharan
Esri Contributor

Can you follow these steps to install the framework?

Install deep learning frameworks for ArcGIS—ArcGIS Pro | Documentation 

The error message states there's a missing dpeendancy. It could very well also be you have installed the relevant python modules to the new cloned enviornment, but its not your active python environment and hence the failures

0 Kudos