mediapipe_model_maker.model_util.LiteRunner

A runner to do inference with the TFLite model.

mediapipe_model_maker.model_util.LiteRunner(
    tflite_model: bytearray
)

tflite_model A valid flatbuffer representing the TFLite model.

Methods

run

View source

run(
    input_tensors: Union[List[tf.Tensor], Dict[str, tf.Tensor]]
) -> Union[List[tf.Tensor], tf.Tensor]

Runs inference with the TFLite model.

Args
input_tensors List / Dict of the input tensors of the TFLite model. The order should be the same as the keras model if it's a list. It also accepts tensor directly if the model has only 1 input.

Returns
List of the output tensors for multi-output models, otherwise just the output tensor. The order should be the same as the keras model.