-
Notifications
You must be signed in to change notification settings - Fork 0
/
wisdm.py
316 lines (267 loc) · 12.3 KB
/
wisdm.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
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
import os
import glob
import numpy as np
import pandas as pd
from os.path import join
from torch.utils.data import Dataset
import torch
from abc import ABC
import configparser
package_dir, _ = os.path.split(os.path.abspath(__file__))
config = configparser.ConfigParser()
for fName in ['config_local.ini', 'config.ini']:
f_path = join(package_dir, fName)
if os.path.isfile(f_path):
config.read(f_path)
continue
class WISDM(Dataset, ABC):
"""Abstract class for the WISDM datasets accessible with pytorch"""
def __init__(self,
accel_files,
gyro_files,
root,
time_window=5,
overwrite=False,
folds=[1, 2, 3, 4, 5],
normalize_acc=False):
"""
Args:
root (string): Root directory of dataset where directory ``wisdm`` exists
accel_files (string): name of the accelartion file
gyro_files (string): name of the gyroscope file
folds (array, int): folds to call for.
time_window (int): The time window of input (unit: second).
normalize_acc (bool, optional): if true then normalize accleration data by 9.8 (gravity offset)
overwrite (bool): overwrite existing npz file
"""
# Dataset generation
self.sample_rate = 20 # 20Hz
self.input_size = int(self.sample_rate * time_window)
self.root = root
self.accel_files = accel_files
self.gyro_files = gyro_files
self.accel_file_paths = glob.glob(join(root, accel_files))
self.gyro_file_paths = glob.glob(join(root, gyro_files))
self.accel_file_paths.sort()
self.gyro_file_paths.sort()
# Process root and related values
self.db_path = join(root, 'raw/watch/{}_{}.npz'.format(
type(self).__name__, self.input_size))
# Read the csv and merge df with folds
if 'df' not in self.__dir__():
dfs = []
for index, (accel_file, gyro_file) in enumerate(zip(self.accel_file_paths, self.gyro_file_paths)):
accel_df = pd.read_csv(accel_file, names=['userID', 'label', 'timestamp', 'accel_x', 'accel_y', 'accel_z'], header=0)
gyro_df = pd.read_csv(gyro_file, names=['timestamp', 'gyro_x', 'gyro_y', 'gyro_z'], header=0)
df = pd.merge(accel_df, gyro_df,on="timestamp")
df['accel_z'] = df['accel_z'].str.replace(';','').astype(float)
df['gyro_z'] = df['gyro_z'].str.replace(';','').astype(float)
if index < 1/5 * len(self.accel_file_paths): df['fold'] = 1 # data from every 1/5 users are marked as a folder
elif index < 2/5 * len(self.accel_file_paths): df['fold'] = 2
elif index < 3/5 * len(self.accel_file_paths): df['fold'] = 3
elif index < 4/5 * len(self.accel_file_paths): df['fold'] = 4
else: df['fold'] = 5
dfs.append(df)
self.df = pd.concat(dfs, axis=0, ignore_index=True)
self.classes = sorted(list(set(self.df['label'])))
self.nClasses = len(self.classes)
# Maybe create dataset
if not os.path.isfile(self.db_path) or overwrite:
self._create_dataset()
# Get item processing
self.folds = folds
self.normalize_acc = normalize_acc
self.get_db_folds() # Fils self.accels and self.labels
def __len__(self):
return len(self.imus)
def get_db_folds(self):
full_db = np.load(self.db_path, allow_pickle=True)
self.imus = []
self.labels = []
self.folds_nb = []
for fold in self.folds:
fold_name = 'fold{}'.format(fold)
print('loading ', fold_name)
imus = full_db[fold_name].item()['imus']
labels = full_db[fold_name].item()['labels']
self.imus.extend(imus)
self.labels.extend(labels)
self.folds_nb.extend([fold, ] * len(labels))
self.imus = [self.preprocess(torch.tensor(i, dtype=torch.float)) for i in self.imus] # PyTorch expects the input tensor and model parameters to have the same dtype, since the model parameters are initialized as FloatTensors, we need to change the input to torch.float.
self.labels = [torch.tensor(l).int() for l in self.labels]
def preprocess(self, imu):
# normalize accel
# imu[:3] *= 9.80665 # TODO: change this to np.max(), by changeing self.accels and self.labels into tensors (instead of list of tensors)
if self.normalize_acc:
imu[:3] = imu[:3] / 9.80665 # gravity offset
return imu
def get_label_mapping(self):
classes = self.classes
mapping = {}
for c in classes:
id = classes.index(c)
mapping[c] = id
return mapping
def __getitem__(self, idx):
imu = self.imus[idx]
label = self.labels[idx]
return imu, label
def _create_dataset(self):
df = self.df
input_size = self.input_size
# Create npz file
print('Creating corresponding npz file...')
dataset = {}
for fold in range(1, 6):
fold_name = 'fold{}'.format(fold)
dataset[fold_name] = {}
dataset[fold_name]['imus'] = []
dataset[fold_name]['labels'] = []
pairs = set(zip(df.userID, df.label))
for pair in pairs:
# each activity is collected for about 3 min = 3x60x20 = 3600 frames
df_temp = df[(df['userID'] == pair[0]) & (df['label'] == pair[1])]
num_sequence = len(df_temp) // input_size
for i in range(num_sequence):
rows = df_temp.iloc[i * input_size : (i + 1) * input_size]
imu = rows[['accel_x','accel_y','accel_z','gyro_x','gyro_y','gyro_z']].values.T
label = self.classes.index(pair[1])
fold = df_temp.iloc[i * input_size].fold
fold_name = 'fold{}'.format(fold)
dataset[fold_name]['imus'].append(imu)
dataset[fold_name]['labels'].append(label)
print('Saving')
np.savez(self.db_path, **dataset)
class WISDMSelect(WISDM):
"""Class for the partial WISDM datasets with selected"""
def __init__(self,
accel_files='raw/watch/accel/*.txt',
gyro_files='raw/watch/gyro/*.txt',
root=config['Paths']['WISDM'],
class_file='activity_key.txt',
class_selected=['walking', 'jogging', 'sitting', 'standing', 'typing', 'teeth', 'pasta', 'drinking'],
time_window=5,
overwrite=False,
folds=[1, 2, 3, 4],
normalize_acc = False):
"""
Args:
root (string): Root directory of dataset where directory ``wisdm`` exists
accel_files (string): name of the accelartion file
gyro_files (string): name of the gyroscope file
class_file (string): name of the activity label file
class_selected (list, string): activities selected for the customized wisdm dataset
folds (array, int): folds to call for.
time_window (int): The time window of input (unit: second).
normalize_acc (bool, optional): if true then normalize accleration data by 9.8 (gravity offset).
overwrite (bool): overwrite existing npz file
"""
# Filter df with selected activities
df_key = pd.read_csv(join(root, class_file), sep=' = ',names=['activity', 'label'], engine='python')
self.label_dict = dict(df_key.values)
self.class_selected = class_selected
label_selected = [self.label_dict[c] for c in class_selected]
accel_file_paths = glob.glob(join(root, accel_files))
gyro_file_paths = glob.glob(join(root, gyro_files))
accel_file_paths.sort()
gyro_file_paths.sort()
dfs = []
for index, (accel_file, gyro_file) in enumerate(zip(accel_file_paths, gyro_file_paths)):
accel_df = pd.read_csv(accel_file, names=['userID', 'label', 'timestamp', 'accel_x', 'accel_y', 'accel_z'], header=0)
gyro_df = pd.read_csv(gyro_file, names=['timestamp', 'gyro_x', 'gyro_y', 'gyro_z'], header=0)
df = pd.merge(accel_df, gyro_df,on="timestamp")
df = df[df['label'].isin(label_selected)]
df['accel_z'] = df['accel_z'].str.replace(';','').astype(float)
df['gyro_z'] = df['gyro_z'].str.replace(';','').astype(float)
if index < 1/5 * len(accel_file_paths): df['fold'] = 1 # data from every 1/5 users are marked as a folder
elif index < 2/5 * len(accel_file_paths): df['fold'] = 2
elif index < 3/5 * len(accel_file_paths): df['fold'] = 3
elif index < 4/5 * len(accel_file_paths): df['fold'] = 4
else: df['fold'] = 5
dfs.append(df)
self.df = pd.concat(dfs, axis=0, ignore_index=True)
super().__init__(
accel_files=accel_files,
gyro_files=gyro_files,
root=root,
time_window=time_window,
overwrite=overwrite,
folds=folds,
normalize_acc=normalize_acc)
def get_label_mapping(self):
label_dict_swap = {letter: name for name, letter in self.label_dict.items()}
classes = self.classes
mapping = {}
for c in classes:
id = classes.index(c)
mapping[label_dict_swap[c]] = id
return mapping
if __name__ == '__main__':
from torch.utils.data import DataLoader
from torch import nn
from torch import optim
time_window = 1
imu_train_set = WISDM(
accel_files='raw/watch/accel/*.txt',
gyro_files='raw/watch/gyro/*.txt',
root=config['Paths']['WISDM'],
time_window=time_window,
overwrite=False,
folds=[1, 2, 3, 4],
)
imu_train_loader = DataLoader(imu_train_set, batch_size=16,
shuffle=True, num_workers=4)
# Load a sample network
net = nn.Sequential(
nn.Conv1d(6, 32, 3, 1), nn.ReLU(), nn.BatchNorm1d(32),
nn.Conv1d(32, 32, 3, 1), nn.ReLU(), nn.BatchNorm1d(32),
nn.Conv1d(32, imu_train_set.nClasses, 5, 1), nn.ReLU(),
nn.AdaptiveAvgPool1d(1)
)
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
imus = imu_train_loader.dataset.imus
labels = imu_train_loader.dataset.labels
# test = np.concatenate([s.unsqueeze(0) for s in imus])
# print(test.shape)
# print(max(np.max(test), -np.min(test)))
# Training loop
# n_epochs = 5
# summary = {'loss': [[] for _ in range(n_epochs)], 'acc': [[] for _ in range(n_epochs)]}
# for e in range(n_epochs):
# for i, (imus, labels) in enumerate(imu_train_loader):
# # Zero the grads
# optimizer.zero_grad()
# # Run the Net
# x = net(imus)
# x = x.view(x.shape[:-1])
# # Optimize nettpr
# loss = criterion(x, labels.long())
# loss.backward()
# optimizer.step()
# summary['loss'][e].append(loss.item())
# # Calculat accuracy
# _, pred = x.data.topk(1, dim=1)
# pred = pred.view(pred.shape[:-1])
# acc = torch.sum(pred == labels)/x.shape[0]
# summary['acc'][e].append(acc.item())
# print('Loss: {}, Accuracy: {}'.format(np.mean(summary['loss'][e]), np.mean(summary['acc'][e])))
# imu_test_set = WISDMSelect(
# folds=[5],
# time_window=time_window,
# overwrite=False)
# imu_test_loader = DataLoader(imu_test_set, batch_size=1,
# shuffle=True, num_workers=4)
# test_accuracy = []
# for i, (imus, labels) in enumerate(imu_test_loader):
# # Run the Net
# x = net(imus)
# x = x.view(x.size()[:-1])
# # loss = criterion(x, labels.long())
# # summary['loss'][e].append(loss.item())
# # Calculat accuracy
# _, pred = x.data.topk(1, dim=1)
# pred = pred.view(pred.shape[:-1])
# acc = torch.sum(pred == labels)/x.shape[0]
# summary['acc'][e].append(acc.item())
# print(np.mean(summary['acc'][e]))