|
1
|
|
|
import os |
|
2
|
|
|
import pickle |
|
3
|
|
|
import logging |
|
4
|
|
|
from fuel.datasets import H5PYDataset |
|
5
|
|
|
|
|
6
|
|
|
logger = logging.getLogger(__name__) |
|
7
|
|
|
|
|
8
|
|
|
|
|
9
|
|
|
class CasasFuel(object): |
|
10
|
|
|
"""CasasFuel Class to retrieve CASAS smart home data as a fuel dataset object |
|
11
|
|
|
|
|
12
|
|
|
Args: |
|
13
|
|
|
dir_name (:obj:`string`): |
|
14
|
|
|
Directory path that contains HDF5 dataset file and complementary dataset information pkl file |
|
15
|
|
|
|
|
16
|
|
|
Attributes: |
|
17
|
|
|
data_filename (:obj:`str`): Path to `data.hdf5` dataset file |
|
18
|
|
|
info (:obj:`dict`): complementary dataset information stored in dict format |
|
19
|
|
|
keys of info includes: |
|
20
|
|
|
|
|
21
|
|
|
""" |
|
22
|
|
|
def __init__(self, dir_name): |
|
23
|
|
|
logger.debug('Load Casas H5PYDataset from ' + dir_name) |
|
24
|
|
|
self.data_filename = dir_name + '/data.hdf5' |
|
25
|
|
|
if os.path.isfile(dir_name + '/info.pkl'): |
|
26
|
|
|
f = open(dir_name + '/info.pkl', 'rb') |
|
27
|
|
|
self.info = pickle.load(f) |
|
28
|
|
|
f.close() |
|
29
|
|
|
|
|
30
|
|
|
def get_dataset(self, which_sets, load_in_memory=False, **kwargs): |
|
31
|
|
|
"""Return fuel dataset object specified by which_sets tuple and load it in memory |
|
32
|
|
|
|
|
33
|
|
|
Args: |
|
34
|
|
|
which_sets (:obj:`tuple` of :obj:`str`): containing the name of splits to load. |
|
35
|
|
|
Valid value are determined by the ``info.pkl`` loaded. |
|
36
|
|
|
You can get the list of split set names by :meth:`get_set_list()`. |
|
37
|
|
|
Usually, if the dataset is split by weeks, the split name is in the form of ``week <num>``. |
|
38
|
|
|
If the dataset is split by days, the split name is in the form of ``day <num>``. |
|
39
|
|
|
load_in_memory (:obj:`bool`, Optional): Default to False. |
|
40
|
|
|
Whether to load the data in main memory. |
|
41
|
|
|
|
|
42
|
|
|
Returns: |
|
43
|
|
|
:class:`fuel.datasets.base.Dataset`: A Fuel dataset object created by |
|
44
|
|
|
:class:`fuel.datasets.h5py.H5PYDataset` |
|
45
|
|
|
""" |
|
46
|
|
|
# Check if sets exist as split name in metadata |
|
47
|
|
|
for set_name in which_sets: |
|
48
|
|
|
if set_name not in self.info['split_sets']: |
|
49
|
|
|
logger.error('set %s not found in splits' % set_name) |
|
50
|
|
|
# Load specified splits and return |
|
51
|
|
|
return H5PYDataset(file_or_path=self.data_filename, |
|
52
|
|
|
which_sets=which_sets, |
|
53
|
|
|
load_in_memory=load_in_memory, **kwargs) |
|
54
|
|
|
|
|
55
|
|
|
def get_set_list(self): |
|
56
|
|
|
"""Get the split set list |
|
57
|
|
|
|
|
58
|
|
|
Returns: |
|
59
|
|
|
:obj:`tuple` of :obj:`str`: A list of split set names |
|
60
|
|
|
""" |
|
61
|
|
|
return self.info['split_sets'] |
|
62
|
|
|
|
|
63
|
|
|
def get_input_dims(self): |
|
64
|
|
|
"""Get the dimension of features |
|
65
|
|
|
|
|
66
|
|
|
Returns: |
|
67
|
|
|
:obj:`int` : the input feature length |
|
68
|
|
|
""" |
|
69
|
|
|
dims = len(self.info['index_to_feature']) |
|
70
|
|
|
return dims |
|
71
|
|
|
|
|
72
|
|
|
def get_output_dims(self): |
|
73
|
|
|
"""Get the dimension of target indices |
|
74
|
|
|
|
|
75
|
|
|
Returns: |
|
76
|
|
|
:obj:`int` : the target indices |
|
77
|
|
|
""" |
|
78
|
|
|
dims = len(self.info['index_to_activity']) |
|
79
|
|
|
return dims |
|
80
|
|
|
|
TinghuiWang /
pyActLearn
