Contents

init(validation:batchSize:maximumIterations:predictionWindowSize:augmentationOptions:algorithm:targetFrameRate:)

Creates a new set of training parameters for an action classifier with the validation dataset.

Declaration

init(validation: MLActionClassifier.ModelParameters.ValidationData = .split(strategy: .automatic), batchSize: Int = MLActionClassifier.__Defaults.batchSize, maximumIterations: Int = MLActionClassifier.__Defaults.maximumIterations, predictionWindowSize: Int = MLActionClassifier.__Defaults.predictionWindowSize, augmentationOptions: MLActionClassifier.VideoAugmentationOptions = [.horizontalFlip], algorithm: MLActionClassifier.ModelParameters.ModelAlgorithmType = .stgcn, targetFrameRate: Double = MLActionClassifier.__Defaults.targetFrameRate)

Discussion

  • validation: A validation dataset represented by an MLActionClassifier.ModelParameters.ValidationData instance.

  • batchSize: The number of videos the training session uses for each of its training iterations.

  • maximumIterations: The largest number of training iterations the training session can use.

  • predictionWindowSize: The number of frames the training session uses to train an action classifier. For example, set to 60 to capture actions that take 2 seconds from videos that have a frame rate of 30 frames per second.

  • augmentationOptions: The variations the training session uses to generate more variety in the training dataset.

  • algorithm: The algorithm the training session uses to train the action classifier.

  • targetFrameRate: The number of frames the training session uses per second of video to train an action classifier.