ONetRunner

class plugins.extract.detect.mtcnn.ONetRunner(weights_path: str, device: device, input_size: int, threshold: float)

Bases: object

Keras O-Net model for MTCNN

Parameters:

• weights_path (str) – The path to the torch model file

• device (torch.device) – The device to run inference on

• input_size (int) – The input size of the model

• threshold (float) – Threshold for O-Net

Methods Summary

__call__(images, rectangle_batch)

Third stage - further refinement and facial landmarks positions with o-net

Methods Documentation

__call__(images: ndarray, rectangle_batch: list[ndarray]) → list[tuple[ndarray, ndarray]]

Third stage - further refinement and facial landmarks positions with o-net

Parameters:

• images (ndarray) – The batch of images to detect faces in

• rectangle_batch (list[ndarray]) – List of numpy.ndarray face candidates from R-Net

Return type:

List of refined final candidates, scores and landmark points from O-Net

__call__(images: ndarray, rectangle_batch: list[ndarray]) → list[tuple[ndarray, ndarray]]

Third stage - further refinement and facial landmarks positions with o-net

Parameters:

• images (ndarray) – The batch of images to detect faces in

• rectangle_batch (list[ndarray]) – List of numpy.ndarray face candidates from R-Net

Return type:

List of refined final candidates, scores and landmark points from O-Net