What should l use feature class or classify raster when export training data for deep learning?

372
6
11-01-2021 01:56 AM
Labels (2)
FrankLee1
New Contributor II

As the subject said,What should l use feature class or classify raster when export training data for deep learning? Do input feature class or classify raster  matter with model types ?

For instance,classify raster is used for classify pixels mission with U-Net or HED model.

 

Tags (2)
0 Kudos
6 Replies
DanPatterson
MVP Esteemed Contributor

Introduction to deep learning—ArcGIS Pro | Documentation

the help topics beginning with this in the help topic tree answers basic questions before you get to specifics for particular aspects


... sort of retired...
0 Kudos
FrankLee1
New Contributor II

Thanks for reply.

l mean for the parameter in the screenshot, feature class or classified raster, how could l choose?The help about this tool is not specific,there is no distinction as  how to choose.

FrankLee1_3-1635920058488.png

 

 

 

0 Kudos
DanPatterson
MVP Esteemed Contributor

Export Training Data For Deep Learning (Image Analyst)—ArcGIS Pro | Documentation

It depends on what you have.  Do you have the required inputs in both raster and vector?


... sort of retired...
0 Kudos
FrankLee1
New Contributor II

sure,l have required inputs in both raster and vector. l make a test, when raster as input, it could only export  data with classified tiles format.

 

0 Kudos
DanPatterson
MVP Esteemed Contributor

Then you are back to my first posts link where you now have what you need to...

If you have existing labeled vector or raster data, you can use the Export Training Data For Deep Learning geoprocessing tool to generate the training data needed for the next step.

So you are good to go.... unless you were expecting something completely different


... sort of retired...
0 Kudos
FrankLee1
New Contributor II

Thanks Dan ,l figure out it

0 Kudos