Using Deep Learning Studio

[ModyBuchbinder]

I have some issues with the studio.

After I created samples (I could not find the way to import polygon layer so I created them in the studio) I try to create chips.

The export option gave me a list of my Enterprise data Stores. So I created a new Data Store that is a folder and gave it as target to the export.

The Export finished successfully.

In the folder I created I can see the image chips as small image files.

Then I try to register this folder and got a message that there are no image chips in the directory.

I tries to go to the model and select this folder but, again, I got a message that it is empty.

What is the correct workflow to creating chips and use it for the model?

Thanks

[JayChen]

Hi ModyBuchbinder,

The way to import training samples in Deep Learning Studio is available under the Collect Training Samples dropdown list within the Prepare training data task (11.4/11.5), or through the Import training samples link within Prepare training data task (11.3 and earlier).

For the image chips you exported within the DLS app, it should show up as an entry within the Manage image chips table, and doesn't need to be registered manually.

When you collect samples and export image chips, and after that you want to use the image chips for training within the same project (each project has a definition of project type being one of the three), and if you see the empty message when selecting a set of image chips you exported within the same project, that sounds like a bug. If you can provide screenshot we can take a closer look.

In general, here are the regular workflows:

1. Create project and go to prepare training data task

2. Configure the project with imagery layer or data store folder, and define the work units

3. optionally, if you have existing training samples, you can import it (with the option to decide whether the imported sampels need reveiw or not)

4. Go through each work unit and complete training sample collection

5. Go through each work unit and approve/reject samples (with the option to approve all)

6. Export image chips to the default raster store (with the option to change to a different data store if desired)

7. Go to Train model task

8. Train a new iteration, and pick the set of image chips you exported at step 6

Which version of ArcGIS Enterprise are you using?