create static method

Future<SelfieSegmentation> create({ SegmentationConfig config = const SegmentationConfig(), })

Creates and initializes a selfie segmentation model instance.

This factory method loads the selected segmentation TensorFlow Lite model from package assets and prepares it for inference.

config: Configuration for model selection, delegates, and output limits. The SegmentationConfig.model field selects which model variant to load.

Returns a fully initialized SelfieSegmentation instance.

Throws SegmentationException if:

• Model file cannot be loaded (SegmentationError.modelNotFound)
• Interpreter creation fails (SegmentationError.interpreterCreationFailed)
• Model validation fails (SegmentationError.unexpectedTensorShape)

Example:

// Default: fast binary model (~244KB)
final segmenter = await SelfieSegmentation.create();

// Multiclass model with per-class masks (~16MB)
final segmenter = await SelfieSegmentation.create(
  config: SegmentationConfig(model: SegmentationModel.multiclass),
);

Implementation

static Future<SelfieSegmentation> create({
  SegmentationConfig config = const SegmentationConfig(),
}) async {
  final effectiveModel = config.model;
  final modelFile = segmentationModelFile(effectiveModel);

  final obj = await _createWithLoader(
    config: config,
    loadInterpreter: (options) => Interpreter.fromAsset(
      'packages/face_detection_tflite/assets/models/$modelFile',
      options: options,
    ),
    loadErrorCode: SegmentationError.modelNotFound,
    loadErrorPrefix: 'Failed to load model $modelFile',
  );
  await obj._initializeTensors();
  return obj;
}