forked from resemble-ai/Resemblyzer
-
Notifications
You must be signed in to change notification settings - Fork 0
/
demo_utils.py
201 lines (166 loc) · 6.63 KB
/
demo_utils.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
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
from mpl_toolkits.axes_grid1 import make_axes_locatable
from matplotlib.animation import FuncAnimation
from resemblyzer import sampling_rate
from matplotlib import cm
from time import sleep, perf_counter as timer
from umap import UMAP
from sys import stderr
import matplotlib.pyplot as plt
import numpy as np
_default_colors = plt.rcParams["axes.prop_cycle"].by_key()["color"]
_my_colors = np.array([
[0, 127, 70],
[255, 0, 0],
[255, 217, 38],
[0, 135, 255],
[165, 0, 165],
[255, 167, 255],
[97, 142, 151],
[0, 255, 255],
[255, 96, 38],
[142, 76, 0],
[33, 0, 127],
[0, 0, 0],
[183, 183, 183],
[76, 255, 0],
], dtype=np.float) / 255
def play_wav(wav, blocking=True):
try:
import sounddevice as sd
# Small bug with sounddevice.play: the audio is cut 0.5 second too early. We pad it to
# make up for that
wav = np.concatenate((wav, np.zeros(sampling_rate // 2)))
sd.play(wav, sampling_rate, blocking=blocking)
except Exception as e:
print("Failed to play audio: %s" % repr(e))
def plot_similarity_matrix(matrix, labels_a=None, labels_b=None, ax: plt.Axes=None, title=""):
if ax is None:
_, ax = plt.subplots()
fig = plt.gcf()
img = ax.matshow(matrix, extent=(-0.5, matrix.shape[0] - 0.5,
-0.5, matrix.shape[1] - 0.5))
ax.xaxis.set_ticks_position("bottom")
if labels_a is not None:
ax.set_xticks(range(len(labels_a)))
ax.set_xticklabels(labels_a, rotation=90)
if labels_b is not None:
ax.set_yticks(range(len(labels_b)))
ax.set_yticklabels(labels_b[::-1]) # Upper origin -> reverse y axis
ax.set_title(title)
cax = make_axes_locatable(ax).append_axes("right", size="5%", pad=0.15)
fig.colorbar(img, cax=cax, ticks=np.linspace(0.4, 1, 7))
img.set_clim(0.4, 1)
img.set_cmap("inferno")
return ax
def plot_histograms(all_samples, ax=None, names=None, title=""):
"""
Plots (possibly) overlapping histograms and their median
"""
if ax is None:
_, ax = plt.subplots()
for samples, color, name in zip(all_samples, _default_colors, names):
ax.hist(samples, density=True, color=color + "80", label=name)
ax.legend()
ax.set_xlim(0.35, 1)
ax.set_yticks([])
ax.set_title(title)
ylim = ax.get_ylim()
ax.set_ylim(*ylim) # Yeah, I know
for samples, color in zip(all_samples, _default_colors):
median = np.median(samples)
ax.vlines(median, *ylim, color, "dashed")
ax.text(median, ylim[1] * 0.15, "median", rotation=270, color=color)
return ax
def plot_projections(embeds, speakers, ax=None, colors=None, markers=None, legend=True,
title="", **kwargs):
if ax is None:
_, ax = plt.subplots(figsize=(6, 6))
# Compute the 2D projections. You could also project to another number of dimensions (e.g.
# for a 3D plot) or use a different different dimensionality reduction like PCA or TSNE.
reducer = UMAP(**kwargs)
projs = reducer.fit_transform(embeds)
# Draw the projections
speakers = np.array(speakers)
colors = colors or _my_colors
for i, speaker in enumerate(np.unique(speakers)):
speaker_projs = projs[speakers == speaker]
marker = "o" if markers is None else markers[i]
label = speaker if legend else None
ax.scatter(*speaker_projs.T, c=[colors[i]], marker=marker, label=label)
if legend:
ax.legend(title="Speakers", ncol=2)
ax.set_title(title)
ax.set_xticks([])
ax.set_yticks([])
ax.set_aspect("equal")
return projs
def interactive_diarization(similarity_dict, wav, wav_splits, x_crop=5, show_time=False):
fig, ax = plt.subplots()
lines = [ax.plot([], [], label=name)[0] for name in similarity_dict.keys()]
text = ax.text(0, 0, "", fontsize=10)
def init():
ax.set_ylim(0.4, 1)
ax.set_ylabel("Similarity")
if show_time:
ax.set_xlabel("Time (seconds)")
else:
ax.set_xticks([])
ax.set_title("Diarization")
ax.legend(loc="lower right")
return lines + [text]
times = [((s.start + s.stop) / 2) / sampling_rate for s in wav_splits]
rate = 1 / (times[1] - times[0])
crop_range = int(np.round(x_crop * rate))
ticks = np.arange(0, len(wav_splits), rate)
ref_time = timer()
def update(i):
# Crop plot
crop = (max(i - crop_range // 2, 0), i + crop_range // 2)
ax.set_xlim(i - crop_range // 2, crop[1])
if show_time:
crop_ticks = ticks[(crop[0] <= ticks) * (ticks <= crop[1])]
ax.set_xticks(crop_ticks)
ax.set_xticklabels(np.round(crop_ticks / rate).astype(np.int))
# Plot the prediction
similarities = [s[i] for s in similarity_dict.values()]
best = np.argmax(similarities)
name, similarity = list(similarity_dict.keys())[best], similarities[best]
if similarity > 0.75:
message = "Speaker: %s (confident)" % name
color = _default_colors[best]
elif similarity > 0.65:
message = "Speaker: %s (uncertain)" % name
color = _default_colors[best]
else:
message = "Unknown/No speaker"
color = "black"
text.set_text(message)
text.set_c(color)
text.set_position((i, 0.96))
# Plot data
for line, (name, similarities) in zip(lines, similarity_dict.items()):
line.set_data(range(crop[0], i + 1), similarities[crop[0]:i + 1])
# Block to synchronize with the audio (interval is not reliable)
current_time = timer() - ref_time
if current_time < times[i]:
sleep(times[i] - current_time)
elif current_time - 0.2 > times[i]:
print("Animation is delayed further than 200ms!", file=stderr)
return lines + [text]
ani = FuncAnimation(fig, update, frames=len(wav_splits), init_func=init, blit=not show_time,
repeat=False, interval=1)
play_wav(wav, blocking=False)
plt.show()
def plot_embedding_as_heatmap(embed, ax=None, title="", shape=None, color_range=(0, 0.30)):
if ax is None:
_, ax = plt.subplots()
if shape is None:
height = int(np.sqrt(len(embed)))
shape = (height, -1)
embed = embed.reshape(shape)
cmap = cm.get_cmap()
mappable = ax.imshow(embed, cmap=cmap)
cbar = plt.colorbar(mappable, ax=ax, fraction=0.046, pad=0.04)
cbar.set_clim(*color_range)
ax.set_xticks([]), ax.set_yticks([])
ax.set_title(title)