forked from AutoResearch/EEG-GAN
-
Notifications
You must be signed in to change notification settings - Fork 0
/
visualize_main.py
264 lines (230 loc) · 11.6 KB
/
visualize_main.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
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
import sys
import warnings
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import torch
from helpers import system_inputs
from helpers.dataloader import Dataloader
from helpers.visualize_pca import visualization_dim_reduction
from helpers.visualize_spectogram import plot_fft_hist, plot_spectogram
def main():
default_args = system_inputs.parse_arguments(sys.argv, file='visualize_main.py')
print('\n-----------------------------------------')
print("System output:")
print('-----------------------------------------\n')
if default_args['data'] != '' and default_args['model'] != '':
raise ValueError("Please specify only one of the following arguments: data, model")
if default_args['channel_index'][0] > -1 and (default_args['pca'] or default_args['tsne']):
print("Warning: channel_index is set to a specific value, but PCA or t-SNE is enabled.\n"
"PCA and t-SNE are only available for all channels. Ignoring channel_index.")
# throw error if checkpoint but csv-file is specified
if default_args['model'] != '' and not default_args['model'].endswith('.pt'):
raise ValueError("Inconsistent parameter specification. 'model' was specified but no model-file (.pt) was given.")
if default_args['data'] != '' and not default_args['data'].endswith('.csv'):
raise ValueError("Inconsistent parameter specification. 'data' was specified but no csv-file was given.")
# throw warning if checkpoint and conditions are given
if default_args['model'] != '' and default_args['kw_conditions'][0] != '':
warnings.warn("Conditions are given, but model is specified. Given conditions will be ignored and taken from the model file if the model file contains the conditions parameter.")
original_data = None
if default_args['data'] != '':
n_conditions = len(default_args['kw_conditions']) if default_args['kw_conditions'][0] != '' else 0
# load data with DataLoader
dataloader = Dataloader(path=default_args['data'],
norm_data=True,
kw_time=default_args['kw_time'],
kw_conditions=default_args['kw_conditions'],
kw_channel=default_args['kw_channel'],)
data = dataloader.get_data(shuffle=False)[:, n_conditions:].numpy()
conditions = dataloader.get_labels()[:, :, 0].numpy()
random = True
elif default_args['model'] != '':
state_dict = torch.load(default_args['model'], map_location='cpu')
n_conditions = state_dict['configuration']['n_conditions'] if 'n_conditions' in state_dict['configuration'].keys() else 0
data = np.concatenate(state_dict['samples'])
if len(data.shape) == 2:
data = data.reshape((1, data.shape[0], data.shape[1]))
if len(data.shape) == 3:
conditions = data[:, :n_conditions, 0]
data = data[:, n_conditions:]
elif len(data.shape) == 4:
# autoencoder samples are saved as (n_samples, type, sequence_length, n_channels)
# type = 0: original, type = 1: reconstructed
conditions = data[:, 0, :n_conditions, 0]
original_data = data[:, 0, n_conditions:]
data = data[:, 1, n_conditions:]
# set channel_plots to True if original_data was found in samples
if not default_args['channel_plots'] and data.shape[-1] > 1:
default_args['channel_plots'] = True
warnings.warn("Original data was found in checkpoint and data contains more than 1 channel. Setting channel_plots to True to improve the visualization quality.")
else:
raise ValueError(f"Invalid shape of data: {data.shape}")
random = False
else:
raise ValueError("Please specify one of the following arguments: csv, checkpoint")
# set channel index
if default_args['channel_index'][0] == -1:
channel_index = np.arange(data.shape[-1])
else:
channel_index = default_args['channel_index']
# -----------------------------
# Normal curve plotting
# -----------------------------
if default_args['n_samples'] > 0:
print(f"Plotting {default_args['n_samples']} samples...")
# create a normal curve plot
if default_args['n_samples'] > data.shape[0]:
warnings.warn(f"n_samples ({default_args['n_samples']}) is larger than the number of samples ({data.shape[0]}).\n"
f"Plotting all available samples instead.")
default_args['n_samples'] = data.shape[0]
if random:
index = np.random.randint(0, data.shape[0]-1, default_args['n_samples'])
else:
index = np.linspace(0, data.shape[0]-1, default_args['n_samples'], dtype=int)
ncols = 1 if not default_args['channel_plots'] else len(channel_index)
fig, axs = plt.subplots(nrows=default_args['n_samples'], ncols=ncols)
picking_type = 'randomly' if random else 'evenly'
if original_data is not None:
comparison = '; reconstructed (blue) vs original (orange)'
else:
comparison = ''
fig.suptitle(f'{picking_type} picked samples' + comparison)
for irow, i in enumerate(index):
if ncols == 1:
for j in channel_index:
if default_args['n_samples'] == 1:
axs.plot(data[i, :, j])
if original_data is not None:
axs.plot(original_data[i, :, j])
else:
axs[irow].plot(data[i, :, j])
if original_data is not None:
axs[irow].plot(original_data[i, :, j])
else:
for jcol, j in enumerate(channel_index):
if default_args['n_samples'] == 1:
axs[jcol].plot(data[i, :, j])
if original_data is not None:
axs[jcol].plot(original_data[i, :, j])
else:
axs[irow, jcol].plot(data[i, :, j])
if original_data is not None:
axs[irow, jcol].plot(original_data[i, :, j])
plt.show()
# -----------------------------
# Loss plotting
# -----------------------------
try:
if default_args['loss'] and default_args['model'] == '':
raise ValueError("Loss plotting only available for checkpoint and not csv")
elif default_args['loss']:
print("Plotting losses...")
# get all losses from state_dict
for key in state_dict.keys():
if 'loss' in key:
plt.plot(state_dict[key], label=key, marker='.')
plt.title('training losses')
plt.legend()
plt.show()
except ValueError as e:
print(e)
# -----------------------------
# Average plotting
# -----------------------------
if default_args['average']:
if n_conditions == 0:
print("Plotting averaged curves...")
else:
print("Plotting averaged curves over each set of conditions...")
# average over conditions
if n_conditions > 0:
conditions_set = np.unique(conditions, axis=0)
# sort samples by condition sets
averaged_data = []
for i, cond in enumerate(conditions_set):
index_cond = np.where(np.sum(conditions == cond, axis=1) == n_conditions)
averaged_data.append(np.mean(data[index_cond], axis=0))
# average over samples
averaged_data = np.array(averaged_data)
else:
averaged_data = np.mean(data, axis=0).reshape(1, data.shape[1], data.shape[2])
conditions_set = ['']
# plot averaged data
ncols = 1 if not default_args['channel_plots'] else len(channel_index)
nrows = averaged_data.shape[0]
fig, axs = plt.subplots(nrows=nrows, ncols=ncols)
if n_conditions == 0:
fig.suptitle('averaged curves')
else:
fig.suptitle('averaged curves over conditions')
for i, cond in enumerate(conditions_set):
if ncols == 1:
for j in channel_index:
if nrows == 1:
axs.plot(averaged_data[i, :, j])
else:
axs[i].plot(averaged_data[i, :, j])
else:
for jcol, j in enumerate(channel_index):
if nrows == 1:
axs[jcol].plot(averaged_data[i, :, j])
else:
axs[i, jcol].plot(averaged_data[i, :, j])
# set legend at the right hand side of the plot;
# legend carries the condition information
# make graph and legend visible within the figure
if not default_args['channel_plots']:
if nrows == 1:
axs.legend([f'{cond}'], loc='center right', bbox_to_anchor=(1, 0.5))
else:
axs[i].legend([f'{cond}'], loc='center right', bbox_to_anchor=(1, 0.5))
else:
if nrows == 1:
axs[-1].legend([f'{cond}'], loc='center right', bbox_to_anchor=(1, 0.5))
else:
axs[i, -1].legend([f'{cond}'], loc='center right', bbox_to_anchor=(1, 0.5))
plt.show()
# -----------------------------
# PCA and t-SNE plotting
# -----------------------------
if default_args['pca'] or default_args['tsne']:
if original_data is None and default_args['comp_data'] != '':
# load comparison data
dataloader_comp = Dataloader(path=default_args['comp_data'],
norm_data=True,
kw_time=default_args['kw_time'],
kw_conditions=default_args['kw_conditions'],
kw_channel=default_args['kw_channel'], )
original_data = dataloader_comp.get_data(shuffle=False)[:, n_conditions:].numpy()
elif original_data is None and default_args['comp_data'] == '':
raise ValueError("No comparison data found for PCA or t-SNE. Please specify a comparison dataset with the argument 'comp_data'.")
if default_args['pca']:
print("Plotting PCA...")
visualization_dim_reduction(original_data, data, 'pca', False, 'pca_file')
if default_args['tsne']:
print("Plotting t-SNE...")
visualization_dim_reduction(original_data, data, 'tsne', False, 'tsne_file')
# -----------------------------
# Spectogram plotting
# -----------------------------
if default_args['spectogram']:
print("Plotting spectograms...")
if data.shape[-1] > 1:
warnings.warn(f"Spectogram plotting is only available for 1 channel but {data.shape[-1]} channels were given. Plotting only the first channel instead.")
fft_data = data[:, :, 0]
else:
fft_data = data
plot_spectogram(fft_data)
# -----------------------------
# FFT plotting
# -----------------------------
if default_args['fft']:
print("Plotting FFT...")
if data.shape[-1] > 1:
warnings.warn(f"FFT plotting is only available for 1 channel but {data.shape[-1]} channels were given. Plotting only the first channel instead.")
fft_data = data[:, :, 0]
else:
fft_data = data
plot_fft_hist(fft_data)
if __name__ == '__main__':
main()