asteroid.models.dcunet module

class asteroid.models.dcunet.BaseDCUNet(architecture, stft_kernel_size=512, stft_stride=None, masknet_kwargs=None)

Bases: asteroid.models.base_models.BaseEncoderMaskerDecoder

Base class for DCUNet and DCCRNet classes.

Parameters:

• stft_kernel_size (int) – STFT frame length to use
• stft_stride (int, optional) – STFT hop length to use.

masknet_class

alias of asteroid.masknn.convolutional.DCUMaskNet

get_model_args()

Arguments needed to re-instantiate the model.

postprocess_encoded(tf_rep)

Hook to perform transformations on the encoded, time-frequency domain representation (output of the encoder) before encoder activation is applied.

Parameters:tf_rep (Tensor of shape (batch, freq, time)) – Output of the encoder, before encoder activation is applied. Returns:Transformed tf_rep

postprocess_masked(masked_tf_rep)

Hook to perform transformations on the masked time-frequency domain representation (result of masking in the time-frequency domain) before decoding.

Parameters:masked_tf_rep (Tensor of shape (batch, n_src, freq, time)) – Masked time-frequency representation, before decoding. Returns:Transformed masked_tf_rep

class asteroid.models.dcunet.DCUNet(architecture, stft_kernel_size=512, stft_stride=None, masknet_kwargs=None)

Bases: asteroid.models.dcunet.BaseDCUNet

DCUNet as proposed in [1].

Parameters:

• architecture (str) – The architecture to use, any of “DCUNet-10”, “DCUNet-16”, “DCUNet-20”, “Large-DCUNet-20”.
• stft_kernel_size (int) – STFT frame length to use
• stft_stride (int, optional) – STFT hop length to use.

References

[1] : “Phase-aware Speech Enhancement with Deep Complex U-Net”, Hyeong-Seok Choi et al. https://arxiv.org/abs/1903.03107

masknet_class

alias of asteroid.masknn.convolutional.DCUMaskNet