mp.tasks.vision.ImageEmbedder

Class that performs embedding extraction on images.

mp.tasks.vision.ImageEmbedder(
    lib: mp.tasks.vision.drawing_styles.face_landmarker.serial_dispatcher.SerialDispatcher,
    handle: ctypes.c_void_p,
    dispatcher: mp.tasks.vision.drawing_styles.face_landmarker.async_result_dispatcher.AsyncResultDispatcher
)

The API expects a TFLite model with optional, but strongly recommended, TFLite Model Metadata.

(kTfLiteUInt8/kTfLiteFloat32)

  • image input of size [batch x height x width x channels].
  • batch inference is not supported (batch is required to be 1).
  • only RGB inputs are supported (channels is required to be 3).
  • if type is kTfLiteFloat32, NormalizationOptions are required to be attached to the metadata for input normalization.

At least one output tensor with: (kTfLiteUInt8/kTfLiteFloat32)

  • N components corresponding to the N dimensions of the returned feature vector for this output layer.
  • Either 2 or 4 dimensions, i.e. [1 x N] or [1 x 1 x 1 x N].

lib The dispatcher to use for all C function calls.
handle The C pointer to the underlying image embedder.
dispatcher The handler for async results.

Methods

close

View source

close()

Closes the ImageEmbedder.

cosine_similarity

View source

@classmethod
cosine_similarity(
    u: mp.tasks.components.containers.Embedding,
    v: mp.tasks.components.containers.Embedding
) -> float

Utility function to compute cosine similarity between two embedding entries.

May return an InvalidArgumentError if e.g. the feature vectors are of different types (quantized vs. float), have different sizes, or have an L2-norm of 0.

Args
u An embedding entry.
v An embedding entry.

Returns
The cosine similarity for the two embeddings.

Raises
ValueError May return an error if e.g. the feature vectors are of different types (quantized vs. float), have different sizes, or have an L2-norm of 0.

create_from_model_path

View source

@classmethod
create_from_model_path(
    model_path: str
) -> 'ImageEmbedder'

Creates an ImageEmbedder object from a TensorFlow Lite model and the default ImageEmbedderOptions.

Note that the created ImageEmbedder instance is in image mode, for embedding image on single image inputs.

Args
model_path Path to the model.

Returns
ImageEmbedder object that's created from the model file and the default ImageEmbedderOptions.

Raises
ValueError If failed to create ImageEmbedder object from the provided file such as invalid file path.
RuntimeError If other types of error occurred.

create_from_options

View source

@classmethod
create_from_options(
    options: mp.tasks.vision.ImageEmbedderOptions
) -> 'ImageEmbedder'

Creates the ImageEmbedder object from image embedder options.

Args
options Options for the image embedder task.

Returns
ImageEmbedder object that's created from options.

Raises
ValueError If failed to create ImageEmbedder object from ImageEmbedderOptions such as missing the model.
RuntimeError If other types of error occurred.

embed

View source

embed(
    image: mp.Image,
    image_processing_options: Optional[mp.tasks.vision.ImageProcessingOptions] = None
) -> mp.tasks.text.TextEmbedderResult

Performs image embedding extraction on the provided MediaPipe Image.

Extraction is performed on the region of interest specified by the roi argument if provided, or on the entire image otherwise.

Args
image MediaPipe Image.
image_processing_options Options for image processing.

Returns
An embedding result object that contains a list of embeddings.

Raises
ValueError If any of the input arguments is invalid.
RuntimeError If image embedder failed to run.

embed_async

View source

embed_async(
    image: mp.Image,
    timestamp_ms: int,
    image_processing_options: Optional[mp.tasks.vision.ImageProcessingOptions] = None
) -> None

Sends live image data to embedder.

The results will be available via the "result_callback" provided in the ImageEmbedderOptions. Embedding extraction is performed on the region of interested specified by the roi argument if provided, or on the entire image otherwise.

Only use this method when the ImageEmbedder is created with the live stream running mode. The input timestamps should be monotonically increasing for adjacent calls of this method. This method will return immediately after the input image is accepted. The results will be available via the result_callback provided in the ImageEmbedderOptions. The embed_async method is designed to process live stream data such as camera input. To lower the overall latency, image embedder may drop the input images if needed. In other words, it's not guaranteed to have output per input image.

The result_callback provides:

  • An embedding result object that contains a list of embeddings.
  • The input image that the image embedder runs on.
  • The input timestamp in milliseconds.

Args
image MediaPipe Image.
timestamp_ms The timestamp of the input image in milliseconds.
image_processing_options Options for image processing.

Raises
ValueError If the current input timestamp is smaller than what the image embedder has already processed.

embed_for_video

View source

embed_for_video(
    image: mp.Image,
    timestamp_ms: int,
    image_processing_options: Optional[mp.tasks.vision.ImageProcessingOptions] = None
) -> mp.tasks.text.TextEmbedderResult

Performs image embedding extraction on the provided video frames.

Extraction is performed on the region of interested specified by the roi argument if provided, or on the entire image otherwise.

Only use this method when the ImageEmbedder is created with the video running mode. It's required to provide the video frame's timestamp (in milliseconds) along with the video frame. The input timestamps should be monotonically increasing for adjacent calls of this method.

Args
image MediaPipe Image.
timestamp_ms The timestamp of the input video frame in milliseconds.
image_processing_options Options for image processing.

Returns
An embedding result object that contains a list of embeddings.

Raises
ValueError If any of the input arguments is invalid.
RuntimeError If image embedder failed to run.

__enter__

View source

__enter__()

Returns self upon entering the runtime context.

__exit__

View source

__exit__(
    exc_type, exc_value, traceback
)

Shuts down the MediaPipe task instance on exit of the context manager.