Creates a new version of a model and begins training. Models are managed as part of an Amazon Rekognition Custom Labels project. The response from CreateProjectVersion is an Amazon Resource Name (ARN) for the version of the model.

Training uses the training and test datasets associated with the project. For more information, see Creating training and test dataset in the Amazon Rekognition Custom Labels Developer Guide.

You can train a model in a project that doesn't have associated datasets by specifying manifest files in the TrainingData and TestingData fields.

If you open the console after training a model with manifest files, Amazon Rekognition Custom Labels creates the datasets for you using the most recent manifest files. You can no longer train a model version for the project by specifying manifest files.

Instead of training with a project without associated datasets, we recommend that you use the manifest files to create training and test datasets for the project.

Training takes a while to complete. You can get the current status by calling DescribeProjectVersions. Training completed successfully if the value of the Status field is TRAINING_COMPLETED.

If training fails, see Debugging a failed model training in the Amazon Rekognition Custom Labels developer guide.

Once training has successfully completed, call DescribeProjectVersions to get the training results and evaluate the model. For more information, see Improving a trained Amazon Rekognition Custom Labels model in the Amazon Rekognition Custom Labels developers guide.

After evaluating the model, you start the model by calling StartProjectVersion.

This operation requires permissions to perform the rekognition:CreateProjectVersion action.