-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathdataset.py
133 lines (117 loc) · 4.71 KB
/
dataset.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
import numpy as np
from torch.utils.data import Dataset
from torch.utils.data import DataLoader
import utils
def clean_dataset(dataset_path):
dataset = np.load(dataset_path)
for data in dataset:
onset_data = data[:, :]
def calculate_low_high(dataset_path):
"""14, 115"""
dataset = np.load(dataset_path)
print(dataset[:, :, 79])
l = 127
h = 0
num_notes = 0
for data in dataset:
onset_data = data[:, :] == 2
# print(onset_data.shape)
sustain_data = data[:, :] == 1
silence_data = data[:, :] == 0
pr = np.stack([onset_data, sustain_data, silence_data],
axis=2).astype(bool)
# pr: (32, 128, 3)
pr_matrix = utils.piano_roll_to_target(pr)
# piano_grid = utils.target_to_3dtarget(pr_matrix, max_note_count=20,
# max_pitch=127,
# min_pitch=0, pitch_pad_ind=110,
# pitch_sos_ind=128,
# pitch_eos_ind=129)
# try:
low, high, _, nn = utils.get_low_high_dur_count(pr_matrix)
nn = nn.max()
if low < l:
l = low
if high > h:
h = high
if nn > num_notes:
num_notes = nn
# print(low, high)
# except ValueError:
# import matplotlib.pyplot as plt
# plt.figure()
# ax = plt.subplot(131)
# ax.imshow(pr[:, :, 0])
# ax = plt.subplot(132)
# ax.imshow(pr[:, :, 1])
# ax = plt.subplot(133)
# ax.imshow(pr[:, :, 2])
# plt.show()
return l, h, num_notes
class PolyphonicDataset(Dataset):
def __init__(self, filepath, shift_low, shift_high):
super(PolyphonicDataset, self).__init__()
self.filepath = filepath
self.shift_low = shift_low
self.shift_high = shift_high
self.data = np.load(self.filepath)
def __len__(self):
# consider data augmentation here
return len(self.data) * (self.shift_high - self.shift_low + 1)
def __getitem__(self, id):
# separate id into (no, shift) pair
no = id // (self.shift_high - self.shift_low + 1)
shift = id % (self.shift_high - self.shift_low + 1) + self.shift_low
data = self.data[no, :, :]
# perform pitch shifting using np.roll and masking
data = np.roll(data, shift, axis=1)
if shift > 0:
data[:, :shift] = 0
elif shift < 0:
data[:, shift:] = 0
# if you want to convert data into MIDI message,
# insert your converter code after this line.
onset_data = data[:, :] == 2
sustain_data = data[:, :] == 1
silence_data = data[:, :] == 0
pr = np.stack([onset_data, sustain_data, silence_data],
axis=2).astype(bool)
# pr: (32, 128, 3)
pr_matrix = utils.piano_roll_to_target(pr)
# 14 - 115, 14
# 21 - 105， 11
piano_grid = utils.target_to_3dtarget(pr_matrix, max_note_count=16,
max_pitch=128,
min_pitch=0, pitch_pad_ind=130,
pitch_sos_ind=128,
pitch_eos_ind=129)
return piano_grid.astype(np.int64)
def reshape_grid_to_plot(piano_grid):
y = np.copy(piano_grid)
y = piano_grid.transpose([1, 0, 2]).reshape((-1, 6 * 32))
return y
if __name__ == '__main__':
file_path = \
r'D:\working\seq2seq-AccArrangement\poly-vae\dataset\pop909+mlpv_t32_fix1'
# l, h, num_notes = calculate_low_high(file_path + '/pop909+mlpv_t32_val_fix1.npy')
# print(l, h, num_notes)
train_dataset = PolyphonicDataset(file_path + '/pop909+mlpv_t32_train_fix1.npy',
-3,
+3) # DO augment on training dataset!
val_dataset = PolyphonicDataset(file_path + '/pop909+mlpv_t32_val_fix1.npy', 0, 0) # DO NOT augment on validation dataset!
print(len(train_dataset), len(val_dataset))
print(train_dataset[1124].shape)
x = train_dataset[1124]
y = reshape_grid_to_plot(x)
import matplotlib.pyplot as plt
plt.imshow(y)
plt.show()
# for i in range(0, )
# # print(train_dataset[(2, 3, 4, 5)])
# data_loader = DataLoader(train_dataset, batch_size=1000, shuffle=True)
# for i_batch, sample_batched in enumerate(data_loader):
# print(i_batch)
# print(len(sample_batched))
# print(type(sample_batched))
# print(sample_batched.shape)
# print(sample_batched[0].shape)