MTCNN

We are using facenet-pytorch to load pre-trained MTCNN model¹, see https://github.com/timesler/facenet-pytorch.

Pre-trained models

mozuma.models.mtcnn.pretrained.torch_mtcnn

Pre-trained PyTorch's MTCNN face detection module by FaceNet

Parameters:

Name	Type	Description	Default
thresholds	Tuple[float, float, float]	MTCNN threshold hyperparameters	required
image_size	Tuple[int, int]	Image size after pre-preprocessing	required
min_face_size	int	Minimum face size in pixels	required
device	torch.device	Torch device to initialise the model weights	required

Returns:

Type	Description
TorchMTCNNModule	Pre-trained MTCNN model

Base model

The MTCNN model is an implementation of a TorchModel.

mozuma.models.mtcnn.modules.TorchMTCNNModule

MTCNN face detection module

Attributes:

Name	Type	Description
thresholds	Tuple[float, float, float]	MTCNN threshold hyperparameters
image_size	Tuple[int, int]	Image size after pre-preprocessing
min_face_size	int	Minimum face size in pixels
device	torch.device	Torch device to initialise the model weights

Provider store

See the stores documentation for usage.

mozuma.models.mtcnn.stores.FaceNetMTCNNStore

Pre-trained model states by Facenet for MTCNN

These are identified by training_id=facenet.

---

1. K. Zhang, Z. Zhang, Z. Li, and Y. Qiao. Joint face detection and alignment using multitask cascaded convolutional networks. IEEE Signal Processing Letters, 23(10):1499–1503, Oct 2016. doi:10.1109/LSP.2016.2603342. ↩