FaceBlendshapesModel class

Runs the MediaPipe Blendshape V2 model, which classifies a packed set of 146 facial landmarks into 52 expression coefficients (blendshapes).

Unlike the other models in this package, this is not an image model: it never sees pixels. Its input is a [1, 146, 2] tensor of landmark coordinates produced by packBlendshapeInput, and its output is 52 coefficients in [0, 1] (a terminal sigmoid keeps them in range). Because the input is tiny and dense, the model runs in well under a millisecond on CPU, so it is pinned to the CPU on every path (no GPU dispatch payoff).

The underlying TFLite model (face_blendshapes.tflite, MLP-Mixer, float16) is extracted from Google's MediaPipe Face Landmarker task bundle. See the official model card for architecture, training data, and intended use: https://storage.googleapis.com/mediapipe-assets/Model%20Card%20Blendshape%20V2.pdf (local copy: doc/model_cards/blendshape_v2_model_card.pdf)

Used internally by FaceDetector in FaceDetectionMode.full; results are exposed via Face.blendshapes, Face.smilingProbability, and the eye-open probability getters.

Properties

hashCode → int

The hash code for this object.

no setter, inherited

runtimeType → Type

A representation of the runtime type of the object.

no setter, inherited

Methods

call(Float32List packed) → Future<Float32List?>

Runs the model on a packed [1, 146, 2] landmark tensor and returns the 52 blendshape coefficients, each clamped to [0, 1].

dispose() → void

Releases all TensorFlow Lite resources held by this model.

noSuchMethod(Invocation invocation) → dynamic

Invoked when a nonexistent method or property is accessed.

inherited

toString() → String

A string representation of this object.

inherited

Operators

operator ==(Object other) → bool

The equality operator.

inherited

Static Methods

create({InterpreterOptions? options}) → Future<FaceBlendshapesModel>

Creates and initializes the model from package assets.

createCompiledFromBuffer(Uint8List modelBytes) → Future<FaceBlendshapesModel>

Creates the model backed by a LiteRT CompiledModel, pinned to the CPU.

createFromBuffer(Uint8List modelBytes, {InterpreterOptions? options}) → Future<FaceBlendshapesModel>

Creates the model from pre-loaded model bytes.

createFromFile(String modelPath, {InterpreterOptions? options}) → Future<FaceBlendshapesModel>

Creates the model from a custom .tflite file on disk.