-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathwrite_pred_info_to_file.py
190 lines (175 loc) · 7.69 KB
/
write_pred_info_to_file.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
#!/usr/bin/env python
__author__ = 'jesse'
import argparse
import gensim
import numpy as np
import operator
import os
import pickle
import random
from functions import get_data_for_classifier, fit_classifier, get_margin_kappa
def main():
behaviors = ["drop", "grasp", "hold", "lift", "look", "lower", "press", "push"]
modalities = ["audio_ispy", "color", "fpfh", "haptic_ispy"] # "finger",
nb_objects = 32
# Convert flags to local variables.
indir = FLAGS_indir
kernel = FLAGS_kernel
word_embeddings_fn = FLAGS_word_embeddings
outdir = FLAGS_outdir
print "reading in folds, labels, predicates, and features..."
with open(os.path.join(indir, 'labels.pickle'), 'rb') as f:
labels = pickle.load(f)
with open(os.path.join(indir, 'predicates.pickle'), 'rb') as f:
predicates = pickle.load(f)
nb_predicates = len(predicates)
gmm_feature_fn = os.path.join(indir, 'gmm_features.pickle')
if os.path.isfile(gmm_feature_fn):
with open(gmm_feature_fn, 'rb') as f:
object_feats = pickle.load(f)
else:
print "... extracting normalized ispy features (removes unary features)"
object_feats = {}
for oidx in range(nb_objects):
print "...... extracting from object " + str(oidx)
with open(os.path.join(indir, str(oidx) + '.pickle'), 'rb') as f:
d = pickle.load(f)
object_feats[oidx] = {}
for b in behaviors:
object_feats[oidx][b] = {}
for m in modalities:
if m in d[b].keys():
object_feats[oidx][b][m] = d[b][m]
print "...... done"
print "writing normalized features to file..."
with open(gmm_feature_fn, 'wb') as f:
pickle.dump(object_feats, f)
print "... done"
contexts = []
for oidx in object_feats:
contexts = []
for b in object_feats[oidx]:
for m in object_feats[oidx][b]:
contexts.append((b, m))
for tidx in range(len(object_feats[oidx][b][m])):
object_feats[oidx][b][m][tidx] = object_feats[oidx][b][m][tidx][0] # one timestep eaach
nb_contexts = len(contexts)
valid_predicates = [pidx for pidx in range(nb_predicates)
if sum([1 if labels[oidx][pidx] == 1 else 0
for oidx in range(nb_objects)]) >= 1
and sum([1 if labels[oidx][pidx] == 0 else 0
for oidx in range(nb_objects)]) >= 1]
with open(os.path.join(indir, 'behavior_annotations.pickle'), 'rb') as f:
behavior_annotations = pickle.load(f)
with open(os.path.join(indir, 'modality_norms.pickle'), 'rb') as f:
modality_annotations = pickle.load(f)
print "... done"
# Pre-calculate matrix of cosine similarity of word embeddings.
print "pre-calculating word embeddings similarities..."
print "... loading word embeddings"
wvb = True if word_embeddings_fn.split('.')[-1] == 'bin' else False
wv = gensim.models.KeyedVectors.load_word2vec_format(word_embeddings_fn, binary=wvb,
limit=150000)
print "...... done"
print "... calculating similarities"
# If missing, give 1 to self and 0 else; give 0 similarity between in and out.
pred_cosine = [[(1 + wv.similarity(predicates[pidx], predicates[pjdx])) / 2.0
if predicates[pjdx] in wv.vocab else 0
for pjdx in range(nb_predicates)]
if predicates[pidx] in wv.vocab else
[0 if pjdx != pidx else 1 for pjdx in range(nb_predicates)]
for pidx in range(nb_predicates)]
print "...... done"
print "... done"
# Fit SVMs.
print "fitting SVMs..."
kappas = [] # pidx, b, m
num_examples = [] # pidx
for pidx in range(nb_predicates):
if pidx not in valid_predicates:
print "... '" + predicates[pidx] + "' insufficient labels"
kappas.append({b: {m: 0 for _b, m in contexts if _b == b} for b, _ in contexts})
num_examples.append(0)
continue
print "... '" + predicates[pidx] + "' fitting"
train_pairs = [(oidx, labels[oidx][pidx]) for oidx in range(nb_objects)
if labels[oidx][pidx] == 0 or labels[oidx][pidx] == 1]
num_examples.append(len(train_pairs))
pc = {}
pk = {}
for b, m in contexts:
if b not in pc:
pc[b] = {}
pk[b] = {}
pc[b][m] = fit_classifier(kernel, b, m, train_pairs, object_feats)
pk[b][m] = get_margin_kappa(pc[b][m], b, m, train_pairs, object_feats,
xval=train_pairs, kernel=kernel)
s = sum([pk[b][m] for b, m in contexts])
for b, m in contexts:
pk[b][m] = pk[b][m] / float(s) if s > 0 else 1.0 / nb_contexts
kappas.append(pk)
print "... done"
print "writing labels.csv..."
with open(os.path.join(outdir, 'labels.csv'), 'w') as f:
h = ['oidx']
h.extend(predicates)
f.write(','.join(h) + '\n')
for oidx in range(nb_objects):
l = [str(oidx)]
l.extend([str(int((labels[oidx][pidx] - 0.5) * 2)) for pidx in range(nb_predicates)])
f.write(','.join(l) + '\n')
print "... done"
print "writing cosine.csv..."
with open(os.path.join(outdir, 'cosine.csv'), 'w') as f:
h = ['predicate']
h.extend(predicates)
f.write(','.join(h) + '\n')
for pidx in range(nb_predicates):
l = [predicates[pidx]]
l.extend([str(pred_cosine[pidx][pjdx]) for pjdx in range(nb_predicates)])
f.write(','.join(l) + '\n')
print "... done"
print "writing kappas.csv..."
with open(os.path.join(outdir, 'kappas.csv'), 'w') as f:
h = ['predicate']
h.extend([';'.join(c) for c in contexts])
f.write(','.join(h) + '\n')
for pidx in range(nb_predicates):
l = [predicates[pidx]]
l.extend([str(kappas[pidx][c[0]][c[1]]) for c in contexts])
f.write(','.join(l) + '\n')
print "... done"
print "writing behaviors.csv..."
with open(os.path.join(outdir, 'behaviors.csv'), 'w') as f:
h = ['predicate']
h.extend(behaviors)
f.write(','.join(h) + '\n')
for pidx in range(nb_predicates):
l = [predicates[pidx]]
l.extend([str(behavior_annotations[pidx][b]) for b in behaviors])
f.write(','.join(l) + '\n')
print "... done"
print "writing modalities.csv..."
with open(os.path.join(outdir, 'modalities.csv'), 'w') as f:
h = ['predicate']
h.extend(modalities)
f.write(','.join(h) + '\n')
for pidx in range(nb_predicates):
l = [predicates[pidx]]
l.extend([str(modality_annotations[pidx][m] if pidx in modality_annotations else 0) for m in modalities])
f.write(','.join(l) + '\n')
print "... done"
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--indir', type=str, required=True,
help="data directory")
parser.add_argument('--kernel', type=str, required=True,
help="SVM kernel to use (linear, poly, rbf)")
parser.add_argument('--word_embeddings', type=str, required=True,
help="word embeddings binary to use")
parser.add_argument('--outdir', type=str, required=True,
help="directory to write out csvs")
args = parser.parse_args()
for k, v in vars(args).items():
globals()['FLAGS_%s' % k] = v
main()