Creating a Project and Assigning Roles

XNAT uses the concept of a "Project" to collect data and assign user permissions relative to that data. Any standard project in XNAT can be Machine Learning-enabled. However, user permissions in the project relative to Machine Learning are slightly complex, because of the permissions required to use the Container Service. The following table describes what activities are available for which standard project roles in XNAT.

Custom user groups are not well-supported by the Machine Learning demo


ActionSite AdministratorsProject OwnersProject MembersProject Collaborators
Install Training Commands(tick)


Enable Training Commands in a Project(tick)(tick)

Upload a Model(tick)(tick)
Add Segmentation Resource to Model(tick)(tick)(tick)
Create a Model Configuration(tick)(tick)

Run Data Conversion Commands(tick)(tick)(tick)
Create A Dataset(tick)(tick)

Launch Training(tick)(tick)(tick)
View Training in TensorBoard(tick)(tick)(tick)
Promote Training Results to New Model(tick)(tick)(tick)
Download Training Results(tick)(tick)(tick)(tick)
Delete Training Results(tick)(tick)

Delete Model(tick)(tick)

Delete Dataset(tick)(tick)

$label.name