Available with Image Server
Deep Learning Studio organizes deep learning model development and use into projects, which include a project type. Pixel classification, object detection, and object classification can all be created in a Deep Learning Studio project. In ArcGIS Enterprise 11.1, Deep Learning Studio introduced a new portal item type called a Deep Learning Studio project. A Deep Learning Studio project organizes the reference resources that are required to complete a deep learning process in a single location in an organization. The way in which projects are organized presents a focused user experience that can extend across an organization's enterprise, using shared services and resources as part of the Web GIS.
Project configuration sections
Project information for the deep learning analysis is contained in the subsections below.
General
This section contains basic information about the project, is required for all of three steps, and includes the following:
- Project name—A unique identifier for the project.
- Project type—References the type of deep learning analysis that will be used during the Run inference step, which specifies the image chip metadata, options, and supported model backbone that will be used for the image chip export.
Caution:
Once the project type is chosen during project creation, it cannot be changed.
- Tags—Keywords to help find the project in the organization.
- Summary—A short description of the project that describes the analysis being conducted.
Groups and users
The project may require additional team members to complete the collection process. This section of the Deep Learning Studio project controls how project work is shared with existing groups in the organization.
Note:
This section can be modified at any time during project configuration, but the prompt to choose a group only appears if the Prepare training data step is selected and the project is not configured for training.
In this step, the groups in Enterprise can be selected if already created. Group creation is not possible within Deep Learning Studio.
Imagery source
This section indicates the imagery source that is used for the Prepare training data step. Any imagery source that is added to the organization can be accessed here. The spatial extent of the imagery source is used to divide into work units and as the input layer for training sample creation.
Note:
In addition to imagery layers hosted in your organization, Deep Learning Studio now supports imagery layers from cloud store and cloud raster store locations.
Work units
Depending on the imagery source specified in the previous section, the work units are created automatically based on one of three configuration options: a grid system defining task units, custom work units, or individual images. The work units divide the training collection process into smaller units that can be collected, reviewed, and approved separately.