Implementation of image cropper Attributes: image: the input image. landmarks: using landmarks information to crop.

Data processor for model evaluation. Attributes: image_root(str): root directory of test set. image_list_file(str): path of

Factory to produce backbone according the backbone_conf.yaml. Attributes: backbone_type(str): which backbone will produce. ba

Computes and stores the average and current value

Factory to produce backbone according the backbone_conf.yaml. Attributes: backbone_type(str): which backbone will produce. ba

Factory to produce head according to the head_conf.yaml Attributes: head_type(str): which head will be produce. head_param(di

Computes and stores the average and current value

Common feature extractor. Attributes: device(object): device to init model.

MTCNN face detection module. This class loads pretrained P-, R-, and O-nets and returns images cropped to include the face only, given raw in

Common feature extractor. Attributes: device(object): device to init model.

For each positive pair, takes the hardest negative sample (with the greatest triplet loss value) to create a triplet Margin should match the

same as mxnet, normalize into [-1, 1] image = (image - 127.5)/128

Convert ndarrays in sample to Tensors. process only one batch every time

CoordConv layer as in the paper.

Evaluator the trained model online. Attributes: test_set(str): which test set to use, currently surpport LFW, CPLFW, ..., IJB-C i

Implementation of face landmark model handler Attributes: model: the face landmark model. device: use cpu or gpu to process.

Implementation of face detection model handler Attributes: model: the face detection model. device: use cpu or gpu to process.

Load a model by network and weights file. Attributes: model(object): the model definition file.

Load a model by network and weights file. Attributes: model(object): the model definition file.

Computes and stores the average and current value

A named tuple describing a ResNet block.

A named tuple describing a ResNet block.

Add a virtual mask in face. Attributes: uv_face_path(str): the path of uv_face. mask_template_folder(str): the directory whe

Define a traditional face model which contains a backbone and a head. Attributes: backbone(object): the backbone of face model.

Define a traditional face model which contains a backbone and a head. Attributes: backbone(object): the backbone of face model.

Implementation of face recognition model handler Attributes: model: the face recognition model. device: use cpu or gpu to process

Implementation of IJBC test protocal.

Factory to produce head according to the head_conf.yaml Attributes: head_type(str): which head will be produce. head_param(di

Implementation of LFW test protocal. Attributes: data_loader(object): a test data loader. pair_list(list): the pair list give

Mobile Inverted Residual Bottleneck Block. Args: block_args (namedtuple): BlockArgs, defined in utils.py. global_params (namedtup

The factory used to produce different pairs parser for different dataset. Attributes: pairs_file(str): the path of the pairs file that wa

Implementation for "Additive Margin Softmax for Face Verification"

Implementation for "Adaptively Scaling Cosine Logits for Effectively Learning Deep Face Representations".

Implementation for "AdaptiveFace: Adaptive Margin and Sampling for Face Recognition".

The pairs parser for agedb.

The pairs parser for agedb.

Implementation for "ArcFace: Additive Angular Margin Loss for Deep Face Recognition"

Implement of Towards Flops-constrained Face Recognition (https://arxiv.org/pdf/1909.00632.pdf):