Use Deep Learning models

Deep learning models can be used to automate the task of recording observations and extracting geographical features from the Analysis Layer. To use deep learning in ArcGIS Excalibur, log in with a Professional or Professional Plus user type and a role with the Imagery Analysis privilege.

Add a deep learning model to a project

Before running a deep learning model, an Excalibur project must be created and configured to use deep learning. The project must contain an ArcGIS Imagery Layer or ArcGIS Tiled Imagery Layer as the project’s Analysis Layer to use deep learning.

To add a deep learning model to the project do the following:

1. Click More from an open project.
2. Click Project Details .

   Project details opens.

3. Click Deep Learning in menu.
4. Click Add Models .

   Only project owners and administrators will have the option to add deep learning models.

5. Choose one or more of the pre-defined object detection and pixel classification models from the list.
   • TextSAM
   • Car Detection – USA
   • Ship Detection (RGB)
   • Building Footprint Extraction - USA
   • Land Cover Classification (Aerial Imagery)
   • Land Cover Classification (Sentinel-2)
   • Land Cover Classification (Landsat 8)
6. Click Save . Alternatively, click Cancel.

   The selected models are added to the project in the Deep Learning Models list.

7. Click Model Details to learn more about each model, or click Remove to remove a model from the list.
8. Click Open Project to return to the project.

Use deep learning to run models

After the project has been configured with one or more deep learning models in the project details they can now be used in the Deep Learning tool to conduct analysis.

To use Deep Learning do the following:

1. Click Tools in an open project.
2. Click Deep Learning under Analysis Tools.

   Deep Learning panel opens.

3. Select a model to run.

   Only the deep learning models that were added in the project details will appear in the tool. Click the information button next to the model in the tool to view more information about that model.

   Deep Learning panel changes to display parameters for the selected model.

4. Select an Area of Interest to run the model. Choose Draw an area and use the Draw Area sketch tool, or choose Use the view extent and pan or zoom in the Map Panel to the desired extent.
5. Adjust Model Arguments as necessary for the selected model. The type and number of arguments are different for each model.
6. Type a Layer Name for the result layer.
7. Choose an output Folder where the layer will be published.
8. Optionally, click Estimate Credits to estimate how many credits will be consumed when running the model.
9. Click Run Model to run the selected model and arguments. Alternatively, click Reset .

   The model will run, which may take several minutes to complete. If there are any errors running the operation, the error details will be returned in the tool directly from the server.

   Note:

   If you navigate away from the tool while it is running, you will not be notified if the operation succeeds or fails. If the operation succeeds while you remain in the tool, you will see a success notification and the results layer will automatically be added to the map. If you navigate away from the tool and the operation succeeds, the results layer will not be automatically added to the map but can be added to the map through the Add Layers menu.