No, the training all the images and metadata that the system has processed in order to build its models. It's how the system "knows" what "Ramirez from Highlander" means.
