比较与torchaudio.transforms.FrequencyMasking的差异

torchaudio.transforms.FrequencyMasking

class torchaudio.transforms.FrequencyMasking(freq_mask_param: int, iid_masks: bool = False)

更多内容详见torchaudio.transforms.FrequencyMasking。

mindspore.dataset.audio.FrequencyMasking

class mindspore.dataset.audio.FrequencyMasking(iid_masks=False, freq_mask_param=0, mask_start=0, mask_value=0.0)

更多内容详见mindspore.dataset.audio.FrequencyMasking。

差异对比

PyTorch：给音频波形施加频域掩码。

MindSpore：给音频波形施加频域掩码。不支持变化的mask_value取值。

分类	子类	PyTorch	MindSpore	差异
参数	参数1	freq_mask_param	freq_mask_param	-
	参数2	iid_masks	iid_masks	-
	参数3	-	mask_start	添加掩码的起始位置
	参数4	-	mask_value	指定填充掩码值，MindSpore计算时无法再更改

代码示例

import numpy as np

fake_specgram = np.array([[[0.17274511, 0.85174704, 0.07162686, -0.45436913],
                           [-1.0271876, 0.33526883, 1.7413973, 0.12313101]]]).astype(np.float32)

# PyTorch
import torch
import torchaudio.transforms as T
torch.manual_seed(1)

transformer = T.FrequencyMasking(freq_mask_param=2, iid_masks=True)
torch_result = transformer(torch.from_numpy(fake_specgram), mask_value=0.0)
print(torch_result)
# Out: tensor([[[ 0.0000,  0.0000,  0.0000,  0.0000],
#               [-1.0272,  0.3353,  1.7414,  0.1231]]])

# MindSpore
import mindspore as ms
import mindspore.dataset.audio as audio

ms.dataset.config.set_seed(2)
transformer = audio.FrequencyMasking(freq_mask_param=2, iid_masks=True, mask_start=0, mask_value=0.0)
ms_result = transformer(fake_specgram)
print(ms_result)
# Out: [[[ 0.          0.          0.          0.        ]
#        [-1.0271876   0.33526883  1.7413973   0.12313101]]]