Classify Objects Using Deep Learning

[JasperWong]

Hi all, 

my first post here requesting for help with a Deep Learning Tool.

I run ArcGIS Pro 2.8.0 with the latest Python Packages and I am trying to run the Classify Objects using Deep Learning tool on a 8-bit unsigned 3-band TIF Raster file (Band 2,3,4) on a local drive. 

I was relieved that my export training samples and train deep learning model went well without a hitch.

Export Training Data

Train Deep Learning Model

However, trying to run the Classify tool on my raster was not successful. The process would waver between 0-1%. I left it overnight for processing with the same situation where it would be stuck at 0%. Is there a reason why the tool is not running on my raster? Is it because of the background pixels in my raster?

Classify Objects using DL

Appreciate any comments or thoughts on my issue.

Thank you. 

hey mate -@BrendanNewell , tagging you in here as well as you might be able to help 🙂

[JasperWong]

these files below were created while the tool was in process but nothing else eventuated while the process got stuck at 0%

[TimG]

Thoughts/Questions...

• Are you running with a GPU (what is it) and have you set it on the Environmental tab?
• Grab a tile from the images directory and do a test classify on that.
• How large is Extract_Bands_Pleiad_2.tif and is it 8-bit unsigned?

[JasperWong]

Hi Tim, 

thanks for your reply. I ran it on a smaller processing extent and it worked! It appears to only work on areas 1000X700m (approximately 0.7km sq) and not larger areas. My entire raster area is 700km2. 

My GPU is a NVDIA Quadro RTX 4000

[Anonymous User]

Hi @JasperWong,

Can you run it on an area around 10 km2 and check how much time it takes.

Additionally, You can try increasing the batch size. This will help run the tool faster.

Thanks,

Sandeep