Use inferencing tools

Available with Image Server

In Deep Learning Studio, the Run inference step uses the provided deep learning model to detect objects, classify objects, or classify pixels. Depending on the project type, various output options and model parameters are available during the Run inference step.

Note:

For the Run inference step to be completed, a deep learning package or deep learning model must be available to the user in the organization. If neither is available, one must be added to the organization or generated in the Train model step.

Create output from a deep learning model

The first step in the process is to create an iteration.

Note:

To complete this step, sign in to Deep Learning Studio and open a project unless a project is already open.

1. On the What would you like to do page, click Run inference to open the tool.

   The inferencing tool page appears with the input imagery source and the project type indicating the inferencing tool, and the tool parameters are visible.

2. Optionally, modify the inferencing tool parameters.

Several parameters can be modified for the inferencing process. The parameters that should be considered and modified for each processing run are the following:

• Area of interest—The area considered during the process is dependent on the perception of the deep learning model. If the model needs additional work, a smaller area of interest may be appropriate.
• Output feature layer/imagery layer name—Depending on the project type, this parameter can be a feature layer or imagery layer. Provide a name that is meaningful for this attempt. There may be more than one deep learning output layer in the project, which reinforces the importance of a meaningful name.

3. Click Start to run the inferencing tool.
A dialog box appears with messages about the inferencing tool process.
4. When the tool completes, click Dismiss to close the dialog box.
The map now contains the results of the inferencing tool and the output layer is visible in the organization.
5. Review the output of the inferencing tool in the map.
6. If the inferencing tool output is satisfactory, click the Back button twice to return to the project view.

The Run inference step creates an output layer based on the selected inferencing tool. The results layer is available in the organization for use in other ArcGIS applications. If the output is not satisfactory, consider returning to the Prepare training data or Train model step to continue to improve the deep learning model.