fastface

fastface.list_pretrained_models() → List[str]

Returns available pretrained model names

Returns:

list of pretrained model names

Return type:

List[str]

>>> import fastface as ff
>>> ff.list_pretrained_models()
['lffd_original', 'lffd_slim']

fastface.download_pretrained_model(model: str, target_path: str | None = None) → str

Downloads pretrained model to given target path, if target path is None, it will use model cache path. If model already exists in the given target path than it will do notting.

Parameters:

• model (str) – pretrained model name to download

• target_path (str, optional) – target directory to download model. Defaults to None.

Returns:

file path of the model

Return type:

str

fastface.list_archs() → List[str]

Returns available architecture names

Returns:

list of arch names

Return type:

List[str]

>>> import fastface as ff
>>> ff.list_archs()
['lffd']

fastface.list_arch_configs(arch: str) → List[str]

Returns available architecture configurations as list

Parameters:

arch (str) – architecture name

Returns:

list of arch config names

Return type:

List[str]

>>> import fastface as ff
>>> ff.list_arch_configs('lffd')
['original', 'slim']

fastface.get_arch_config(arch: str, config: str) → Dict

Returns configuration dictionary for given arch and config names

Parameters:

• arch (str) – architecture name

• config (str) – configuration name

Returns:

configuration details as dictionary

Return type:

Dict

>>> import fastface as ff
>>> ff.get_arch_config('lffd', 'slim')
{'input_shape': (-1, 3, 480, 480), 'backbone_name': 'lffd-v2', 'head_infeatures': [64, 64, 64, 128, 128], 'head_outfeatures': [128, 128, 128, 128, 128], 'rf_sizes': [20, 40, 80, 160, 320], 'rf_start_offsets': [3, 7, 15, 31, 63], 'rf_strides': [4, 8, 16, 32, 64], 'scales': [(10, 20), (20, 40), (40, 80), (80, 160), (160, 320)]}