-
Notifications
You must be signed in to change notification settings - Fork 0
/
fic.py
143 lines (120 loc) · 4.7 KB
/
fic.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
"""FIC dataset"""
from collections import Counter
import pickle
import csv
import os
import datetime as dt
import logging
import glob
import tensorflow as tf
import numpy as np
FREQUENCY = 100
DEFAULT_LABEL = "Idle"
FLIP_ACC = [-1., 1., 1.]
FLIP_GYRO = [1., -1., -1.]
TIME_FACTOR = 1
TRAIN_IDS = ['1_1','1_2','2_1','2_2','2_3','3_1','3_2','3_3','6_1','6_2','6_3',
'9_1','10_1','11_1']
VALID_IDS = ['4_1','4_2','4_3','7_1']
TEST_IDS = ['5_1','8_1','12_1']
def _int64_feature(value):
return tf.train.Feature(int64_list=tf.train.Int64List(value=[value]))
def _bytes_feature(value):
return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))
def _floats_feature(value):
return tf.train.Feature(float_list=tf.train.FloatList(value=value))
class Dataset():
def __init__(self, src_dir, exp_dir, dom_hand_spec, label_spec,
label_spec_inherit, exp_uniform, exp_format):
self.src_dir = src_dir
self.exp_dir = exp_dir
self.exp_uniform = exp_uniform
self.exp_format = exp_format
# Read data
pickle_path = os.path.join(src_dir, "fic_pickle.pkl")
if not os.path.isfile(pickle_path):
raise RunimeError('Pickle file not found')
with open(pickle_path, 'rb') as f:
self._data = pickle.load(f)
# Class counters
self.counts_1 = Counter()
def __add_to_class_counts(self, class_counts, labels):
"""Add increment to class counts"""
unique, counts = np.unique(labels, return_counts=True)
new_class_counts = Counter(dict(zip(unique, counts)))
return class_counts + new_class_counts
def ids(self):
return [(self._data['subject_id'][i], self._data['session_id'][i]) \
for i in range(0, len(self._data['subject_id']))]
def check(self, id):
return True
def data(self, i, id):
logging.info("Reading processed data from pickle")
timestamps = self._data['signals'][i][:,0]
acc = self._data['signals'][i][:,1:4]
gyro = self._data['signals'][i][:,4:7]
return timestamps, {"hand": (acc, gyro)}
def labels(self, i, id, timestamps):
num = len(timestamps)
# Read annotations
annotations = self._data['bite_gt'][i]
# Read labels
labels_1 = np.empty(num, dtype='U25')
labels_1.fill(DEFAULT_LABEL)
for start_time, end_time in annotations:
start_frame = np.argmax(np.array(timestamps) >= start_time)
end_frame = np.argmax(np.array(timestamps) > end_time)
labels_1[start_frame:end_frame] = "Intake"
# Update class names
self.counts_1 = self.__add_to_class_counts(self.counts_1, labels_1)
return list(labels_1)
def dominant(self, id):
return "NA"
def write(self, path, id, timestamps, data, _, label_1):
frame_ids = list(range(0, len(timestamps)))
id_s = '_'.join([str(x) for x in id])
def _format_time(t):
return (dt.datetime.min + dt.timedelta(seconds=t)).time().strftime('%H:%M:%S.%f')
timestamps = [_format_time(t) for t in timestamps]
acc = np.asarray(data["hand"][0])
gyro = np.asarray(data["hand"][1])
assert len(timestamps) == len(acc), \
"Number timestamps and acc readings must be equal"
assert len(timestamps) == len(gyro), \
"Number timestamps and acc readings must be equal"
if self.exp_format == 'csv':
with open(path, 'w', newline='') as csvfile:
writer = csv.writer(csvfile, delimiter=',')
writer.writerow(["id", "frame_id", "timestamp", "acc_x", "acc_y",
"acc_z", "gyro_x", "gyro_y", "gyro_z", "label_1"])
for i in range(0, len(timestamps)):
writer.writerow([id_s, frame_ids[i], timestamps[i],
acc[i][0], acc[i][1], acc[i][2], gyro[i][0], gyro[i][1],
gyro[i][2], label_1[i]])
elif self.exp_format == 'tfrecord':
with tf.io.TFRecordWriter(path) as tfrecord_writer:
for i in range(0, len(timestamps)):
example = tf.train.Example(features=tf.train.Features(feature={
'example/subject_id': _bytes_feature(id_s.encode()),
'example/frame_id': _int64_feature(frame_ids[i]),
'example/timestamp': _bytes_feature(timestamps[i].encode()),
'example/acc': _floats_feature(acc[i].ravel()),
'example/gyro': _floats_feature(gyro[i].ravel()),
'example/label_1': _bytes_feature(label_1[i].encode())
}))
tfrecord_writer.write(example.SerializeToString())
def done(self):
logging.info("Done")
logging.info("Final number of frames for category 1: {0}.".format(self.counts_1))
def get_flip_signs(self):
return FLIP_ACC, FLIP_GYRO
def get_frequency(self):
return FREQUENCY
def get_time_factor(self):
return TIME_FACTOR
def get_train_ids(self):
return TRAIN_IDS
def get_valid_ids(self):
return VALID_IDS
def get_test_ids(self):
return TEST_IDS