Select to view content in your preferred language

Classify Pixels Using Deep Learning tool outputs an empty layer

334
4
05-23-2024 01:31 AM
Labels (1)
JoštBogataj
New Contributor

Hi,

I am trying do use Esri's Land Cover Classification model on aerial images in ArcGIS version 3.0.2.

Whenever I run the tool the process is completed but the output layer is empty. I have tried running it with different input layers, models and parameters but the result is always the same. Every time the process goes to 2 % and than skips to 'Generating raster' stage. What's weird is that the output layer has 1 row, 1 column and 1 channel. 
Input raster information:

JotBogataj_0-1716453057543.png

 

Output raster information:

JotBogataj_1-1716453057913.png

 

Any help would be appreciated.

0 Kudos
4 Replies
David_McRitchie
Esri Contributor

If I recall correctly I have seen this before when the Deep Learning Essentials installer has not be successfully installed.

 

Mind checking that this is available within your Program and Files? The download link to this can be found here.

 

Hope that helps,

David

Esri UK -Technical Support Analyst
0 Kudos
SuyaZheng
New Contributor

Have you solved the problem? I also encountered this situation, version 3.0.1, and confirmed the following facts:
1. Correctly installed the deep learning package
2. The training model can use pure GPU to obtain results, but inference will result in 2% and then skip outputting an empty layer
3. Model inference using pure CPU can obtain correct results, but it is obvious that the waiting time is too long

0 Kudos
JoštBogataj
New Contributor

Hi,

Have you managed to solve the problem by any chance?

0 Kudos
JoštBogataj
New Contributor

I did not solve the issue even after reinstalling the Deep Learning Essentials package. I got the same results as you, trying CPU only processing.

0 Kudos