CreateML trains and evaluates custom Core ML models directly on device, spanning classifiers, regressors, and image, text, sound, and tabular tasks. You load and shape training data with MLDataTable, MLDataColumn, and MLDataValue, then fit models such as MLImageClassifier, MLSoundClassifier, MLObjectDetector, MLRecommender, MLLinearRegressor, or tree-based estimators like MLBoostedTreeClassifier and MLRandomForestRegressor. Training results are assessed through metrics types including MLClassifierMetrics, MLRegressorMetrics, and MLObjectDetectorMetrics, while MLJob, MLProgress, MLPhase, and MLCheckpoint let you monitor and manage long-running training sessions. Specialized estimators such as MLActionClassifier, MLHandPoseClassifier, MLStyleTransfer, and MLGazetteer cover action recognition, pose, image styling, and named-entity tasks.
Training Data 6
Load, shape, and represent the tabular data you use to train and evaluate models.
- StMLDataTableiOS 15.0+A table of data for training or evaluating a machine learning model.
- StMLDataColumniOS 15.0+A column of typed values in a data table.
- StMLUntypedColumniOS 15.0+A column of untyped values in a data table.
- EnMLDataValueiOS 15.0+The value of a cell in a data table.
- PrMLDataValueConvertibleiOS 15.0+A type that can convert itself to and from a data value.
- EnMLSplitStrategyiOS 15.0+Data partitioning approaches, typically for creating a validation dataset
Image Models 3
Estimators that train Core ML models for image classification, object detection, and image styling.
- StMLImageClassifierA model you train to classify images.
- StMLObjectDetectorA model you train to classify one or more objects within an image.
- StMLStyleTransferA model you train to apply an image’s style to other images or videos.
Sound and Motion Models 5
Estimators that train models recognizing sounds, activities, and hand-based actions or poses.
- StMLSoundClassifierA machine learning model you train with audio files to recognize and identify sounds on a device.
- StMLActionClassifierA model you train with videos to classify a person’s body movements.
- StMLActivityClassifierA model you train to classify motion sensor data.
- StMLHandActionClassifierA task that creates a hand action classification model by training with
- StMLHandPoseClassifierA task that creates a hand pose classification model by training with images
Text and Language Models 4
Estimators that train models for text classification, named-entity tagging, and word representation.
- StMLTextClassifierA model you train to classify natural language text.
- StMLWordTaggerA word-tagging model you train to classify natural language text at the word level.
- StMLWordEmbeddingA map of strings in a vector space that enable your app to find similar
- StMLGazetteerA collection of terms and their labels, which augments a tagger that analyzes natural language text.
Tabular Classifiers and Regressors 11
General-purpose classifier and regressor estimators that learn from tabular feature data.
- StMLBoostedTreeClassifieriOS 15.0+A classifier based on a collection of decision trees combined with gradient boosting.
- StMLBoostedTreeRegressoriOS 15.0+A regressor based on a collection of decision trees combined with gradient boosting.
- StMLDecisionTreeClassifieriOS 15.0+A classifier that predicts the target by creating rules to split the data.
- StMLDecisionTreeRegressoriOS 15.0+A regressor that estimates the target by learning rules to split the data.
- StMLRandomForestClassifieriOS 15.0+A classifier based on a collection of decision trees trained on subsets of the data.
- StMLRandomForestRegressoriOS 15.0+A regressor based on a collection of decision trees trained on subsets of the data.
- StMLLinearRegressoriOS 15.0+A regressor that estimates the target as a linear function of the features.
- StMLLogisticRegressionClassifieriOS 15.0+A classifier that predicts a discrete target value as a function of data features.
- StMLSupportVectorClassifierA classifier that predicts a binary target value by maximizing the separation between categories.
- EnMLClassifierA model you train to classify data into discrete categories.
- EnMLRegressorA model you train to estimate continuous values.
Recommendation 1
An estimator that trains a model to recommend items from interaction data.
- StMLRecommenderA model you train to make recommendations based on item similarity,
Evaluation Metrics 5
Types that report a trained model's accuracy and performance on validation or test data.
- StMLClassifierMetricsiOS 15.0+Metrics you use to evaluate a classifier’s performance.
- StMLRegressorMetricsiOS 15.0+Metrics you use to evaluate a regressor’s performance.
- StMLObjectDetectorMetricsMetrics you use to evaluate an object detector’s performance.
- StMLRecommenderMetricsMetrics you use to evaluate a recommender’s performance.
- StMLWordTaggerMetricsMetrics you use to evaluate a word tagger’s performance.
Training Sessions and Progress 9
Manage, monitor, checkpoint, and visualize long-running asynchronous training sessions.
- ClMLJobiOS 15.0+The representation of a model’s asynchronous training session you use to
- ClMLTrainingSessioniOS 15.0+The current state of a model’s asynchronous training session.
- StMLTrainingSessionParametersiOS 15.0+The configuration settings for a training session.
- StMLCheckpointiOS 15.0+The state of a model’s asynchronous training session at a specific point in
- StMLProgressiOS 15.0+A convenience type that exposes information about the progress of a training
- EnMLPhaseiOS 15.0+The possible states of a training session.
- PrMLVisualizableAn image visualization of machine learning types.
- PrMLStreamingVisualizableA sequence of image visualizations for machine learning types.
- PrMLIdentifierA type the Create ML framework can use as a machine learning identifier.
Model Metadata 1
Author-supplied information attached to a trained Core ML model.
- StMLModelMetadataiOS 15.0+Information about a model that’s stored in a Core ML model file.
Bounding Box Configuration 3
Constants describing how object-detection bounding boxes are anchored, oriented, and measured.
- EnMLBoundingBoxAnchorA location within a bounding box that an annotation’s coordinates use as
- EnMLBoundingBoxCoordinatesOriginThe location within an image that an annotation’s coordinates use as their
- EnMLBoundingBoxUnitsThe units a bounding box annotation uses to define its position and size.
Errors 1
Errors that can occur while creating or training a model.
- EnMLCreateErroriOS 15.0+The errors Create ML throws while performing various operations, such as