-
Notifications
You must be signed in to change notification settings - Fork 0
/
objectome_behavior.py
executable file
·190 lines (155 loc) · 6.84 KB
/
objectome_behavior.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
import random
import time
import os
import pymongo
import scipy
import numpy as np
import cPickle as pk
import tabular as tb
import scipy.io as io
import scipy.stats as stats
import objectome as obj
import dldata.metrics.utils as utils
import dldata.human_data.roschlib as rl
from scipy.stats import norm
from objectome import strip_objectomeFileName, get_index_MWorksRelative
from sklearn.metrics import confusion_matrix
from scipy.stats.mstats import zscore
BKPDATAPATH = obj.dicarlolab_homepath + '/monkey_objectome/human_behaviour/data/mongodb_bkp/'
KW_NAMES = ['sample_obj', 'dist_obj', 'choice', 'id', 'WorkerID', 'AssignmentID']
KW_FORMATS = ['|S40','|S40','|S40','|S40','|S40','|S40']
class psychophysDatasetObject(object):
""" Dataset object for human psychophysics: formats data into tabarray with KW_NAMES entries. """
def __init__(self, collection, selector, meta=None, mongo_reload=False):
self.collection = collection
self.collection_backup = BKPDATAPATH + self.collection + '_bkp.pkl'
self.trial_kw_names = KW_NAMES
self.trial_kw_formats = KW_FORMATS
if mongo_reload | (not os.path.isfile(self.collection_backup)) :
conn = pymongo.Connection(port = 22334, host = 'localhost')
db = conn.mturk
col = db[collection]
data = list(col.find(selector))
print 'Loaded from mongoDB ...'
self.splitby = 'id'
self.data = data
self.meta = meta
self.get_data()
self.backup_to_disk()
else:
self.collection = collection
self.load_from_disk()
print 'Loaded from local mongo backup...'
nsubs = len(np.unique(self.trials['WorkerID']))
print '%d trials from %d subjects' % (self.trials.shape[0], nsubs)
def backup_to_disk(self):
dat = {
'collection':self.collection,
'data':self.data,
'trials':self.trials,
'meta':self.meta,
'splitby':self.splitby
}
with open(self.collection_backup, 'wb') as _f:
pk.dump(dat, _f)
print 'Backed up to ' + self.collection_backup
def load_from_disk(self):
with open(self.collection_backup, 'r') as _f:
dat = pk.load(_f)
self.collection = dat['collection']
self.data = dat['data']
self.trials = dat['trials']
self.meta = dat['meta']
self.splitby = dat['splitby']
return
def get_data(self):
trial_records = []
obj_oi = np.unique(self.meta['obj'])
img_oi = np.unique(self.meta['id'])
for subj in self.data:
for r,i,ss in zip(subj['Response'], subj['ImgData'], subj['StimShown']):
if len(i) > 1:
s = i['Sample']
t = i['Test']
s_id = s['id']
s_obj = s['obj']
t_obj = [t_['obj'] for t_ in t]
d_obj = [t_ for t_ in t_obj if t_ != s_obj][0]
resp = t_obj[r]
else: #backwards compatibility with previous mturkutils
s_id = i[0]['id']
s_obj = i[0]['obj']
t_obj = [strip_objectomeFileName(fn) for fn in ss[1:]]
d_obj = [t_ for t_ in t_obj if t_ != s_obj][0]
resp = strip_objectomeFileName(r)
if (s_id in img_oi) & (d_obj in obj_oi):
rec_curr = (s_obj,) + (d_obj,) + (resp,) + (s_id,) + (subj['WorkerID'],) + (subj['AssignmentID'],)
trial_records.append(rec_curr)
self.trials = tb.tabarray(records=trial_records, names=self.trial_kw_names, formats=self.trial_kw_formats)
return
def get_monkeyturk_data(dataset='objectome24'):
if dataset == 'objectome24':
meta_path = obj.dicarlolab_homepath + 'stimuli/objectome24s100/metadata.pkl'
# data_path = obj.dicarlolab_homepath + 'monkeyturk/allData_v2.mat'
data_path = obj.dicarlolab_homepath + 'monkeyturk/allData_v3.mat'
meta = pk.load(open(meta_path,'r'))
datmat = io.loadmat(data_path)
uobjs = obj.models_combined24
col_data_seg = {}
trial_records = []
subjs = ['Manto', 'Zico', 'Picasso', 'Nano', 'Bento', 'Magneto', 'Pablo']
for sub in subjs:
x = datmat['allData'][sub][0,0]
for xi in range(x.shape[0]):
s_obj = uobjs[x[xi,0]]
d_obj = uobjs[x[xi,2]]
resp = uobjs[x[xi,3]]
s_id = meta[x[xi,4]-1]['id']
workid = sub
assnid = 'MonkeyTurk'
rec_curr = (s_obj,) + (d_obj,) + (resp,) + (s_id,) + (workid,) + (assnid,)
trial_records.append(rec_curr)
col_data_seg['pool'] = tb.tabarray(records=trial_records, names=KW_NAMES, formats=KW_FORMATS)
for sub in subjs:
t = col_data_seg['pool']['WorkerID'] == sub
col_data_seg[sub] = col_data_seg['pool'][t]
return col_data_seg
def get_model_data(dataset='objectome24'):
if dataset == 'objectome24':
featurespath = obj.dicarlolab_homepath + 'stimuli/objectome24s100/features/'
meta = obj.objectome24_meta()
all_metas, all_features = {}, {}
# f_oi = ['ALEXNET_fc6', 'ALEXNET_fc8', 'RESNET101_conv5', 'VGG_fc6', 'VGG_fc8', 'ALEXNET_fc7', 'GOOGLENET_pool5', 'V1', 'VGG_fc7']
f_oi = ['RESNET101_conv5']
for f in f_oi:
data = np.load(featurespath + f + '.npy')
all_features[f] = data
all_metas[f] = meta
return obj.testFeatures(all_features, all_metas, f_oi, obj.models_combined24)
# def get_neural_data(dataset='objectome24'):
def composite_dataset(dataset='objectome24', meta=None, threshold=12000, mongo_reload=False):
if dataset == 'objectome24':
collections = ['objectome64', 'objectome_imglvl', 'ko_obj24_basic_2ways', 'monkobjectome', 'ko_obj24_basic_2ways_mod_ver2']
if meta == None:
meta = obj.objectome24_meta()
elif dataset == 'hvm10':
collections = ['hvm10_basic_2ways', 'hvm10_allvar_basic_2ways'] #, 'hvm10_basic_2ways_newobj', 'hvm10-finegrain']
if meta == None:
meta = obj.hvm10_meta()
fns = ['sample_obj', 'id', 'dist_obj', 'choice', 'WorkerID']
col_data = ()
for col in collections:
dset = obj.psychophysDatasetObject(col, {}, meta, mongo_reload=mongo_reload)
col_data = col_data + (dset.trials,)
trials = tb.rowstack(col_data)
# segregate into pool and individuals
workers = trials['WorkerID']
col_data_seg = {'all':trials, 'pool':()}
for uw in np.unique(workers):
tw = np.nonzero([w == uw for w in workers])[0]
if len(tw) < threshold:
col_data_seg['pool'] = col_data_seg['pool'] + (trials[tw],)
else:
col_data_seg[uw] = trials[tw]
col_data_seg['pool'] = tb.rowstack(col_data_seg['pool'])
return col_data_seg