-
Notifications
You must be signed in to change notification settings - Fork 6
/
learn-features.py
executable file
·172 lines (155 loc) · 7.21 KB
/
learn-features.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
#!/bin/ipython
# ---LICENSE-BEGIN - DO NOT CHANGE OR MOVE THIS HEADER
# This file is part of the Neurorobotics Platform software
# Copyright (C) 2014,2015,2016,2017 Human Brain Project
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
# ---LICENSE-END
import numpy as np
import cv2
import sys
import pyNN.nest as sim
import pathlib as plb
import time
import pickle
import argparse as ap
import re
import common as cm
import network as nw
import visualization as vis
import time
parser = ap.ArgumentParser('./c1-spikes-from-file-test.py --')
parser.add_argument('--c1-dumpfile', type=str, required=True,
help='The output file to contain the C1 spiketrains')
parser.add_argument('--epoch-size', type=int, default=30,
help='The lenght of an epoch')
parser.add_argument('--feature-size', type=int, default=3,
help='The size of the features to be learnt')
parser.add_argument('--s2-prototype-cells', type=int, default=3,
help='The number of S2 features to compute')
parser.add_argument('--plot-c1-spikes', action='store_true',
help='Plot the spike trains of the C1 layers')
parser.add_argument('--plot-s2-spikes', action='store_true',
help='Plot the spike trains of the S2 layers')
#parser.add_argument('--refrac-s2', type=float, default=.1, metavar='.1',
# help='The refractory period of neurons in the S2 layer in\
# ms')
parser.add_argument('--threads', default=1, type=int)
args = parser.parse_args()
sim.setup(threads=args.threads, min_delay=.1)
layer_collection = {}
# Read the gabor features for reconstruction
feature_imgs_dict = {} # feature string -> image
for filepath in plb.Path('features_gabor').iterdir():
feature_imgs_dict[filepath.stem] = cv2.imread(filepath.as_posix(),
cv2.CV_8UC1)
# Extracting meta-information about the simulation from the filename
c1_dumpfile_name = plb.Path(args.c1_dumpfile).stem
image_count = int(re.search('\d*imgs', c1_dumpfile_name).group()[:-4])
sim_time = float(re.search('\d+\.\d+ms', c1_dumpfile_name).group()[:-2])
dataset_label = '{}_fs{}_{}prots'.format(c1_dumpfile_name, args.feature_size,
args.s2_prototype_cells)
print('Create C1 layers')
t1 = time.clock()
dumpfile = open(args.c1_dumpfile, 'rb')
ddict = pickle.load(dumpfile)
layer_collection['C1'] = {}
for size, layers_as_dicts in ddict.items():
layer_list = []
for layer_as_dict in layers_as_dicts:
n, m = layer_as_dict['shape']
spiketrains = layer_as_dict['segment'].spiketrains
dimensionless_sts = [[s for s in st] for st in spiketrains]
new_layer = nw.Layer(sim.Population(n * m,
sim.SpikeSourceArray(spike_times=dimensionless_sts),
label=layer_as_dict['label']), (n, m))
layer_list.append(new_layer)
layer_collection['C1'][size] = layer_list
print('C1 creation took {} s'.format(time.clock() - t1))
print('Creating S2 layers')
t1 = time.clock()
layer_collection['S2'] = nw.create_S2_layers(layer_collection['C1'],
args.feature_size,
args.s2_prototype_cells,
refrac_s2=6)
print('S2 creation took {} s'.format(time.clock() - t1))
#for layers in layer_collection['C1'].values():
# for layer in layers:
# layer.population.record('spikes')
#for layer_list in layer_collection['S2'].values():
# for layer in layer_list:
# layer.population.record(['spikes', 'v'])
reconstructions_dir_dataset_path = plb.Path('S2_reconstructions/' + dataset_label)
if not reconstructions_dir_dataset_path.exists():
reconstructions_dir_dataset_path.mkdir(parents=True)
if args.plot_c1_spikes:
c1_plots_dir_path = plb.Path('plots/C1/' + dataset_label)
if not c1_plots_dir_path.exists():
c1_plots_dir_path.mkdir(parents=True)
if args.plot_s2_spikes:
s2_plots_dataset_dir = plb.Path('plots/S2/' + dataset_label)
for i in range(args.s2_prototype_cells):
s2_plots_dir_path = s2_plots_dataset_dir / str(i)
if not s2_plots_dir_path.exists():
s2_plots_dir_path.mkdir(parents=True)
epoch_weights = [] # type: List[Tuple[int, List[Dict[str, np.array]]]]
# Let the simulation run to "fill" the layer pipeline with spikes
sim.run(40)
print('========= Start simulation =========')
start_time = time.clock()
for i in range(image_count):
print('Simulating for image number', i)
sim.run(sim_time)
if args.plot_c1_spikes:
vis.plot_C1_spikes(layer_collection['C1'],
'{}_image_{}'.format(dataset_label, i),
out_dir_name=c1_plots_dir_path.as_posix())
if args.plot_s2_spikes:
vis.plot_S2_spikes(layer_collection['S2'],
'{}_image_{}'.format(dataset_label, i),
args.s2_prototype_cells,
out_dir_name=s2_plots_dataset_dir.as_posix())
if (i + 1) % 10 == 0:
current_weights = nw.get_current_weights(layer_collection['S2'],
args.s2_prototype_cells)
cv2.imwrite('{}/{}_{:0>4}_images.png'.format(\
reconstructions_dir_dataset_path.as_posix(),
dataset_label, i + 1),
vis.reconstruct_S2_features(current_weights,
feature_imgs_dict,
args.feature_size))
if (i + 1) % args.epoch_size == 0:
current_weights = nw.get_current_weights(layer_collection['S2'],
args.s2_prototype_cells)
epoch_weights.append((i + 1, current_weights))
end_time = time.clock()
print('========= Stop simulation =========')
print('Simulation took', end_time - start_time, 's')
# Reconstruct the last image
cv2.imwrite('{}/{}_{:0>4}_images.png'.format(\
reconstructions_dir_dataset_path.as_posix(),
dataset_label, i + 1),
vis.reconstruct_S2_features(current_weights,
feature_imgs_dict,
args.feature_size))
# Also add the weights of the last iteration to the dumpfile
if image_count % args.epoch_size != 0:
epoch_weights.append((image_count, current_weights))
dumpfile_name = 'S2_weights/{}.bin'.format(dataset_label)
out_dumpfile = open(dumpfile_name, 'wb')
print('Dumping weights for the selected epochs to file', dumpfile_name)
pickle.dump(epoch_weights, out_dumpfile, protocol=4)
out_dumpfile.close()
sim.end()