-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathutils.py
More file actions
168 lines (142 loc) · 6.56 KB
/
utils.py
File metadata and controls
168 lines (142 loc) · 6.56 KB
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
164
165
166
167
168
import os
import os.path
import torch
import sys
import torchaudio
from torchvision.utils import save_image
import librosa
import soundfile as sf
from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image
class SaveUtils():
def __init__(self, args, name):
self.args = args
self.save_dir = os.path.join(args.saveDir, name) # make
if not os.path.exists(self.save_dir):
os.makedirs(self.save_dir)
self.save_dir_image = os.path.join(self.save_dir, 'IMAGE')
if not os.path.exists(self.save_dir_image):
os.makedirs(self.save_dir_image)
self.save_dir_lms = os.path.join(self.save_dir, 'LMS')
if not os.path.exists(self.save_dir_lms):
os.makedirs(self.save_dir_lms)
self.save_dir_model = os.path.join(self.save_dir, 'model')
if not os.path.exists(self.save_dir_model):
os.makedirs(self.save_dir_model)
if os.path.exists(self.save_dir + '/log.txt'):
self.logFile = open(self.save_dir + '/log.txt', 'a')
else:
self.logFile = open(self.save_dir + '/log.txt', 'w')
def save_log(self, log):
sys.stdout.flush()
self.logFile.write(log + '\n')
self.logFile.flush()
def save_model(self, model, epoch):
torch.save(model.state_dict(), self.save_dir_model + '/model_'+ str(epoch) + '.pt')
def save_image(self, gt, fake, epoch):
save_image(gt, self.save_dir_image +'/gt_img_'+ str(epoch) +'.png')
save_image(fake, self.save_dir_image +'/output_img_'+ str(epoch) +'.png')
def save_image_onlyGT(self, gt, epoch):
save_image(gt, self.save_dir_image +'/gt_img_'+ str(epoch) +'.png')
def save_mel(self, gt, fake, epoch, label):
cmap = plt.get_cmap('jet')
instruments = ['bassoon', 'cello', 'clarinet', 'double_bass', 'flute', 'horn', 'oboe', 'sax', 'trombone', 'trumpet', 'tuba', 'viola', 'violin']
batch_size = gt.shape[0]
for i in range(batch_size):
plt.figure(figsize=(2,4))
plt.matshow(gt[i], cmap=cmap)
plt.clim(-100, 50)
plt.axis('off')
plt.title(instruments[label[i]], fontsize=25)
plt.savefig(self.save_dir_lms +'/'+ str(i)+'gt.png')
gt_images = [Image.open(self.save_dir_lms +'/'+ str(x)+'gt.png') for x in range(batch_size)]
widths, heights = zip(*(i.size for i in gt_images))
total_width = int(sum(widths)/2) + widths[0]
max_height = int(max(heights)*2)
new_im = Image.new('RGB', (total_width, max_height))
x_offset = 0
i=0
for im in gt_images:
if i>=int(batch_size/2)+1:
new_im.paste(im, (x_offset,im.size[1]))
x_offset += im.size[0]
else:
new_im.paste(im, (x_offset,0))
x_offset += im.size[0]
if i==int(batch_size/2):
x_offset=0
i+=1
new_im.save(self.save_dir_lms +'/gt_mel_'+ str(epoch) +'.jpg')
del gt_images
del new_im
for i in range(batch_size):
os.remove(self.save_dir_lms +'/'+ str(i)+'gt.png')
for i in range(batch_size):
plt.figure(figsize=(2,4))
plt.matshow(fake[i], cmap=cmap)
plt.clim(-100, 50)
plt.axis('off')
plt.title(instruments[label[i]], fontsize=25)
plt.savefig(self.save_dir_lms +'/'+ str(i)+'fake.png')
fake_images = [Image.open(self.save_dir_lms +'/'+ str(x)+'fake.png') for x in range(batch_size)]
widths, heights = zip(*(i.size for i in fake_images))
total_width = int(sum(widths)/2) + widths[0]
max_height = int(max(heights)*2)
new_im = Image.new('RGB', (total_width, max_height))
x_offset = 0
i=0
for im in fake_images:
if i>=int(batch_size/2)+1:
new_im.paste(im, (x_offset,im.size[1]))
x_offset += im.size[0]
else:
new_im.paste(im, (x_offset,0))
x_offset += im.size[0]
if i==int(batch_size/2):
x_offset=0
i+=1
new_im.save(self.save_dir_lms +'/output_mel_'+ str(epoch) +'.jpg')
del fake_images
del new_im
for i in range(batch_size):
os.remove(self.save_dir_lms +'/'+ str(i)+'fake.png')
def save_mel_onlyGT(self, gt, epoch, label):
cmap = plt.get_cmap('jet')
instruments = ['bassoon', 'cello', 'clarinet', 'double_bass', 'flute', 'horn', 'oboe', 'sax', 'trombone', 'trumpet', 'tuba', 'viola', 'violin']
batch_size = gt.shape[0]
for i in range(batch_size):
plt.figure(figsize=(2,4))
plt.matshow(gt[i], cmap=cmap)
plt.clim(-100, 50)
plt.axis('off')
plt.title(instruments[label[i]], fontsize=25)
plt.savefig(self.save_dir_lms +'/'+ str(i)+'gt.png')
gt_images = [Image.open(self.save_dir_lms +'/'+ str(x)+'gt.png') for x in range(batch_size)]
widths, heights = zip(*(i.size for i in gt_images))
total_width = int(sum(widths)/2) + widths[0]
max_height = int(max(heights)*2)
new_im = Image.new('RGB', (total_width, max_height))
x_offset = 0
i=0
for im in gt_images:
if i>=int(batch_size/2)+1:
new_im.paste(im, (x_offset,im.size[1]))
x_offset += im.size[0]
else:
new_im.paste(im, (x_offset,0))
x_offset += im.size[0]
if i==int(batch_size/2):
x_offset=0
i+=1
new_im.save(self.save_dir_lms +'/gt_mel_'+ str(epoch) +'.jpg')
del gt_images
'''
def save_audio(self, fake, epoch, label):
instruments = ['bassoon', 'cello', 'clarinet', 'double_bass', 'flute', 'horn', 'oboe', 'sax', 'trombone', 'trumpet', 'tuba', 'viola', 'violin']
audio = 10.0**((fake[0]/10))
audio = librosa.feature.inverse.mel_to_audio(audio, sr=44100, n_fft=2048, hop_length=512)
print(audio)
sf.write(self.save_dir_lms +'/'+ instruments[label[0]]+"(output_lms_reconstructed).wav", audio, 44100)
'''