featureExtractor
The underlying base model the training session uses to extract image features as it trains an image classifier.
Declaration
var featureExtractor: MLImageClassifier.FeatureExtractorType { get set }Discussion
When you train an image classifier with Create ML, you don’t train a complete model that converts images to labels. Instead, Create ML leverages a feature extractor, which is another model that identifies a set of general but distinguishing image characteristics. You can either use Create ML’s scenePrint feature extractor or provide your own custom feature extractor.
In either case, you train your model to map the feature extractor’s output to labels in a process known as transfer learning. Your model effectively borrows the knowledge of the feature extractor to accelerate its own training process while requiring fewer training images.