-
Notifications
You must be signed in to change notification settings - Fork 66
/
Copy pathloss.py
163 lines (126 loc) · 6.38 KB
/
loss.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
# Copyright (c) 2021 Sony Corporation. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
'''
MSE and SDR Combination loss definition for MSS
'''
import nnabla.functions as F
def mse(x, y):
# l2 distance and reduce mean
se = F.squared_error(x, y)
return F.mean(se)
def unsqueeze(x):
# unsqueeze at first axis
return F.reshape(x, (1,) + x.shape)
def mse_loss(mix_spec, msk_hat, gt_spec):
# MSE-Combination Loss
# mix_spec -> (Fbin, BatchSize(16), 2(channels), Frames)
# msk_hat -> (Fbin, BatchSize(16), 8(4source x 2channels), Frames)
# channel-dim -> [bass1, bass2, drums1, drums2, ...]
# gt_spec -> (Fbin, BatchSize(16), 8(4source x 2channels), Frames)
assert msk_hat.shape == gt_spec.shape
m_1 = msk_hat[Ellipsis, 0:2, :] # bass
m_2 = msk_hat[Ellipsis, 2:4, :] # drums
m_3 = msk_hat[Ellipsis, 4:6, :] # vocals
m_4 = msk_hat[Ellipsis, 6:8, :] # other
gt_1 = gt_spec[Ellipsis, 0:2, :] # bass
gt_2 = gt_spec[Ellipsis, 2:4, :] # drums
gt_3 = gt_spec[Ellipsis, 4:6, :] # vocals
gt_4 = gt_spec[Ellipsis, 6:8, :] # other
# 4C1 Combination Losses
loss_mse_1 = mse(m_1*mix_spec, gt_1)
loss_mse_2 = mse(m_2*mix_spec, gt_2)
loss_mse_3 = mse(m_3*mix_spec, gt_3)
loss_mse_4 = mse(m_4*mix_spec, gt_4)
# 4C2 Combination Losses
loss_mse_5 = mse((m_1+m_2)*mix_spec, (gt_1+gt_2))
loss_mse_6 = mse((m_1+m_3)*mix_spec, (gt_1+gt_3))
loss_mse_7 = mse((m_1+m_4)*mix_spec, (gt_1+gt_4))
loss_mse_8 = mse((m_2+m_3)*mix_spec, (gt_2+gt_3))
loss_mse_9 = mse((m_2+m_4)*mix_spec, (gt_2+gt_4))
loss_mse_10 = mse((m_3+m_4)*mix_spec, (gt_3+gt_4))
# 4C3 Combination Losses
loss_mse_11 = mse((m_1+m_2+m_3)*mix_spec, (gt_1+gt_2+gt_3))
loss_mse_12 = mse((m_1+m_2+m_4)*mix_spec, (gt_1+gt_2+gt_4))
loss_mse_13 = mse((m_1+m_3+m_4)*mix_spec, (gt_1+gt_3+gt_4))
loss_mse_14 = mse((m_2+m_3+m_4)*mix_spec, (gt_2+gt_3+gt_4))
# All 14 Combination Losses (4C1 + 4C2 + 4C3)
loss_mse = (loss_mse_1 + loss_mse_2 + loss_mse_3 + loss_mse_4 + loss_mse_5 + loss_mse_6 + loss_mse_7 +
loss_mse_8 + loss_mse_9 + loss_mse_10 + loss_mse_11 + loss_mse_12 + loss_mse_13 + loss_mse_14) / 14.0
return loss_mse
def sdr_loss(mix, pred, gt_time):
# SDR-Combination Loss
# mix -> (BatchSize(16), 2(1 source x 2 channels), TimeLen) -> (B, C, T)
# pred -> (4(sources), Bsize, 2(channels), Len) -> (S, B, C, T)
# gt_time -> (BatchSize(16), 8(4 source x 2 channels), TimeLen) -> (B, S*C, T)
# channel-dim -> [bass1, bass2, drums1, drums2, ...]
_, batch_size, n_channels, length = pred.shape
# Fix Length
mix = mix[Ellipsis, :length]
gt_time = gt_time[Ellipsis, :length]
# Fix Shape
mix = unsqueeze(mix) # [1, B, C, T]
gt_time = unsqueeze(gt_time) # [1, B, S*C, T]
data_t = mix # [1, B, C, T]
for i in range(4):
data_t = F.concatenate(data_t, gt_time[Ellipsis, 2*i:2*i+2, :], axis=0)
data_t = F.reshape(data_t, (-1, length)) # [5*B*C, T]
pred = F.reshape(pred, (batch_size*n_channels *
pred.shape[0], pred.shape[-1])) # [B*C*S, T]
# Combination List (4C2 + 4C3)
combi_list = [(1, 2), (1, 3), (1, 4), (2, 3), (2, 4), (3, 4),
(1, 2, 3), (1, 2, 4), (1, 3, 4), (2, 3, 4)]
for combi in combi_list:
if len(combi) == 2:
tmp_data = data_t[batch_size*n_channels*combi[0]:batch_size*n_channels*(
combi[0]+1), Ellipsis] + data_t[batch_size*n_channels*combi[1]:batch_size*n_channels*(combi[1]+1), Ellipsis]
tmp_pred = pred[batch_size*n_channels*(combi[0]-1):batch_size*n_channels*combi[0], Ellipsis] + \
pred[batch_size*n_channels*(
combi[1]-1):batch_size*n_channels*combi[1], Ellipsis]
else:
tmp_data = data_t[batch_size*n_channels*combi[0]:batch_size*n_channels*(combi[0]+1), Ellipsis] + data_t[batch_size*n_channels*combi[1]:batch_size*n_channels*(
combi[1]+1), Ellipsis] + data_t[batch_size*n_channels*combi[2]:batch_size*n_channels*(combi[2]+1), Ellipsis]
tmp_pred = pred[batch_size*n_channels*(combi[0]-1):batch_size*n_channels*combi[0], Ellipsis] + pred[batch_size*n_channels*(
combi[1]-1):batch_size*n_channels*combi[1], Ellipsis] + pred[batch_size*n_channels*(combi[2]-1):batch_size*n_channels*combi[2], Ellipsis]
data_t = F.concatenate(data_t, tmp_data, axis=0)
pred = F.concatenate(pred, tmp_pred, axis=0)
# All 14 Combinations (4C1 + 4C2 + 4C3)
mix_t = F.tile(data_t[:batch_size*n_channels, Ellipsis], (14, 1))
data_t = data_t[batch_size*n_channels:, Ellipsis]
# SDR Loss Calculation
loss_sdr = sdr_loss_core(pred, data_t, mix_t, weighted=True)
return 1.0 + loss_sdr
def sdr_loss_core(inp, gt, mix, weighted=True):
assert inp.shape == gt.shape # (Batch, Len)
assert mix.shape == gt.shape # (Batch, Len)
inp = inp[:, 200:-200]
gt = gt[:, 200:-200]
mix = mix[:, 200:-200]
ns = mix - gt
ns_hat = mix - inp
if weighted:
alpha = F.sum((gt*gt), 1, keepdims=True) / (F.sum((gt*gt), 1,
keepdims=True) + F.sum((ns*ns), 1, keepdims=True) + 1e-10)
else:
alpha = 0.5
# Target
num_cln = F.sum((inp*gt), 1, keepdims=True)
denom_cln = ((1e-10 + F.sum((inp*inp), 1, keepdims=True))
** 0.5) * ((1e-10 + F.sum((gt*gt), 1, keepdims=True)) ** 0.5)
sdr_cln = num_cln / (denom_cln + 1e-10)
# Noise
num_noise = F.sum((ns*ns_hat), 1, keepdims=True)
denom_noise = ((1e-10 + F.sum((ns_hat*ns_hat), 1, keepdims=True))
** 0.5) * ((1e-10 + F.sum((ns*ns), 1, keepdims=True)) ** 0.5)
sdr_noise = num_noise / (denom_noise + 1e-10)
return F.mean(-alpha*sdr_cln - (1. - alpha)*sdr_noise)