-
Notifications
You must be signed in to change notification settings - Fork 8
/
common_jupyter.py
160 lines (125 loc) · 6.33 KB
/
common_jupyter.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
import evaluation
import pandas as pd
import os
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import datasets
import re
from glob import glob
import tensorflow as tf
from tensorboard.backend.event_processing.event_accumulator import EventAccumulator
def get_param_num(results_dir):
tensorboard_dir = results_dir + "/../"
event_acc = EventAccumulator(tensorboard_dir)
event_acc.Reload()
tensor_event = event_acc.Tensors("model_info")
tensor_np = tf.make_ndarray(tensor_event[0].tensor_proto)
return int(tensor_np[0, 1])
def load_data(paths, attributes=None, attr_names=None):
data = []
for attrs, path in zip(attributes, paths):
results = evaluation.results("test", path, ".csv")
if attributes:
for name, attr in zip(attr_names, attrs):
results[name] = attr
data.append(results)
return pd.concat(data)
def ld(attr_names, attr_types, regex, experiments_dir, verbose=False, results_dir="model-f0-outputs"):
experiments_paths = get_paths(experiments_dir, results_dir)
paths = list(filter(lambda x: re.search(regex, x), experiments_paths))
assert len(paths) != 0
attributes = get_attrs_from_paths(regex, attr_types, paths)
if verbose:
a = map(str, zip([p.split("/")[-2] for p in paths], attributes))
print("\n".join(a))
return load_data(paths, attributes, attr_names)
def get_attrs_from_paths(regex, types, paths):
out = []
#for attr_re, type_fn in zip(attr_re_list, types):
for path in paths:
values = re.findall(regex, path)
assert len(values) == 1
values = values[0]
if not isinstance(values, tuple):
values = (values,)
out.append([type_fn(v) for v, type_fn in zip(values, types)])
return out
def to_latex(df):
df = df.reset_index()
# print(df.columns.values.tolist())
aliases = {"Raw Pitch Accuracy": "RPA", "Raw Chroma Accuracy": "RCA", "Voicing Accuracy": "VA", "Overall Accuracy": "OA", "Voicing False Alarm": "VFA", "Voicing Recall": "VR"}
header = map(lambda x: aliases[x] if x in aliases else x, df.columns.values)
print(df.to_latex(float_format=lambda x: "%.3f"%x, bold_rows=True, header=list(header), index=False))
def get_min_max(datas, attr_names, split="MedleyDB valid."):
means = []
for d in datas:
means.append(d[d.Dataset == split].groupby(attr_names).mean())
means = pd.concat(means)
vminRPA, vmaxRPA = means["Raw Pitch Accuracy"].min(), means["Raw Pitch Accuracy"].max()
vminRCA, vmaxRCA = means["Raw Chroma Accuracy"].min(), means["Raw Chroma Accuracy"].max()
return vminRPA, vmaxRPA, vminRCA, vmaxRCA
def plot_grid(data, attr_names, name="test", split="MedleyDB valid.", axs=None, vminRPA=None, vmaxRPA=None, vminRCA=None, vmaxRCA=None):
if axs is None:
sns.set(rc={'figure.figsize': (10, 4)})
sns.set(style="whitegrid")
fig, axs = plt.subplots(ncols=2, sharex=True, sharey=True)
axs[1].set_ylabel("")
plt.tight_layout()
cmap = sns.cubehelix_palette(100, reverse=True, as_cmap=True)
pivot = data[data.Dataset==split].groupby(attr_names).mean().reset_index().pivot(attr_names[0], attr_names[1], "Raw Pitch Accuracy")
sns.heatmap(pivot, annot=True, fmt=".3f", cmap=cmap, ax=axs[0], vmin=vminRPA, vmax=vmaxRPA)
pivot = data[data.Dataset==split].groupby(attr_names).mean().reset_index().pivot(attr_names[0], attr_names[1], "Raw Chroma Accuracy")
sns.heatmap(pivot, annot=True, fmt=".3f", cmap=cmap, ax=axs[1], vmin=vminRCA, vmax=vmaxRCA)
if axs is None:
fig.savefig("figures/"+name+".pdf", bbox_inches="tight")
return pivot
def plot_data(data, attr_names, name="test", split="MedleyDB valid.", plot_metric="Raw Pitch Accuracy", palette="cubehelix", drop_metrics=["Voicing Accuracy",'Voicing Recall', 'Voicing False Alarm', "Overall Accuracy"], figsize=(8, None), order=None):
if split is not None:
data = data[data.Dataset==split]
hue = None
# palette = sns.cubehelix_palette(8)
num_bars = len(data.groupby(attr_names))
categories = data[attr_names[0]].unique()
if len(attr_names) > 1:
hue = attr_names[1]
categories = data[attr_names[1]].unique()
# palette = sns.cubehelix_palette(8)
if palette == "cubehelix":
palette = sns.cubehelix_palette(len(categories)+2)
if figsize[1] is None:
figsize = (figsize[0], num_bars*0.5)
sns.set(rc={'figure.figsize': figsize})
sns.set(style="whitegrid")
_order = None
if order:
_order = data.groupby(attr_names).mean().reset_index().sort_values(plot_metric)[attr_names[0]]
ax = sns.boxplot(x=plot_metric, y=attr_names[0], orient="h", hue=hue, data=data, fliersize=2, palette=palette, showmeans=True, showfliers=False,
meanprops={"markerfacecolor": "black", "markeredgecolor": "black"}, order=_order)
# sns.swarmplot(x=plot_metric, y=attr_names[0], orient="h", hue=hue, data=data, dodge=True, linewidth=1, edgecolor='gray', palette=palette, alpha=0.7, size=4)
figure = ax.get_figure()
figure.savefig("figures/"+name+".pdf", bbox_inches="tight")
summary = data.drop(drop_metrics, axis=1).groupby(attr_names).mean()
if order:
summary = summary.sort_values(plot_metric)
return summary
def plot_note_hist(method, path, path2, est_suffix=".csv"):
sns.set(rc={'figure.figsize': (8, 6)})
sns.set(style="whitegrid")
diffs = []
for prefix, split, dataset_name, ref_paths, est_paths in evaluation.paths_iterator(method, path, est_suffix):
for filename, (ref_time, ref_freq, est_time, est_freq) in evaluation.load_melody_paths(ref_paths, est_paths):
ref_notes = datasets.common.hz_to_midi_safe(ref_freq)
est_notes = datasets.common.hz_to_midi_safe(np.abs(est_freq))
diff = (est_notes - ref_notes)[(ref_freq > 0) & (est_freq > 0)]
filtered_diff = diff[np.abs(diff) > 0.5]
diffs.append(filtered_diff)
fig, ax = plt.subplots()
bins = np.arange(-5, 5)
ax.set_ylim(0, 50000)
ax.set_xticks(bins)
bins = bins - 0.5
sns.distplot(np.concatenate(diffs), kde=False, bins=bins)
def get_paths(path, results_dir="model-f0-outputs"):
paths = sorted(glob(path+"/*/"+results_dir) + glob(path+"/*/*/"+results_dir))
return list(filter(lambda x: "koš" not in x, paths)) # vyhoď modely v koši