load.py 13.7 KB
Newer Older
André Anjos's avatar
André Anjos committed
1 2 3 4 5 6 7 8 9
#!/usr/bin/env python
# vim: set fileencoding=utf-8 :
# Andre Anjos <andre.anjos@idiap.ch>
# Mon 23 May 2011 16:23:05 CEST

"""A set of utilities to load score files with different formats.
"""

import numpy
10 11 12
import tarfile
import os

13 14 15
import logging
logger = logging.getLogger('bob.measure')

16
def open_file(filename, mode='rt'):
17 18 19
  """open_file(filename) -> file_like

  Opens the given score file for reading.
20 21
  Score files might be raw text files, or a tar-file including a single score file inside.

22
  **Parameters:**
23

24
  ``filename`` : str or file-like
25 26
    The name of the score file to open, or a file-like object open for reading.
    If a file name is given, the according file might be a raw text file or a (compressed) tar file containing a raw text file.
27

28 29 30
  **Returns:**

  ``file_like`` : file-like
31 32
    A read-only file-like object as it would be returned by open().
  """
33 34 35 36
  if not isinstance(filename, str) and hasattr(filename, 'read'):
    # It seems that this is an open file
    return filename

37 38 39
  if not os.path.isfile(filename):
    raise IOError("Score file '%s' does not exist." % filename)
  if not tarfile.is_tarfile(filename):
40
    return open(filename, mode)
41 42 43 44 45 46 47 48 49 50 51 52 53 54

  # open the tar file for reading
  tar = tarfile.open(filename, 'r')
  # get the first file in the tar file
  tar_info = tar.next()
  while tar_info is not None and not tar_info.isfile():
    tar_info = tar.next()
  # check that one file was found in the archive
  if tar_info is None:
    raise IOError("The given file is a .tar file, but it does not contain any file.")

  # open the file for reading
  return tar.extractfile(tar_info)

André Anjos's avatar
André Anjos committed
55 56

def four_column(filename):
57 58 59 60 61 62 63 64 65 66 67 68
  """four_column(filename) -> claimed_id, real_id, test_label, score

  Loads a score set from a single file and yield its lines (to avoid loading the score file at once into memory).
  This function verifies that all fields are correctly placed and contain valid fields.
  The score file must contain the following information in each line:

    claimed_id real_id test_label score

  **Parametes:**

  ``filename`` : str or file-like
    The file object that will be opened with :py:func:`open_file` containing the scores.
André Anjos's avatar
André Anjos committed
69

70
  **Yields:**
André Anjos's avatar
André Anjos committed
71

72 73
  ``claimed_id`` : str
    The claimed identity -- the client name of the model that was used in the comparison
André Anjos's avatar
André Anjos committed
74

75 76 77 78 79 80 81 82
  ``real_id`` : str
    The real identity -- the client name of the probe that was used in the comparison

  ``test_label`` : str
    A label of the probe -- usually the probe file name, or the probe id

  ``score`` : float
    The result of the comparison of the model and the probe
André Anjos's avatar
André Anjos committed
83 84
  """

85 86
  for i, l in enumerate(open_file(filename)):
    if isinstance(l, bytes): l = l.decode('utf-8')
André Anjos's avatar
André Anjos committed
87 88 89 90 91 92 93 94 95
    s = l.strip()
    if len(s) == 0 or s[0] == '#': continue #empty or comment
    field = [k.strip() for k in s.split()]
    if len(field) < 4:
      raise SyntaxError('Line %d of file "%s" is invalid: %s' % (i, filename, l))
    try:
      score = float(field[3])
    except:
      raise SyntaxError('Cannot convert score to float at line %d of file "%s": %s' % (i, filename, l))
96
    yield (field[0], field[1], field[2], score)
André Anjos's avatar
André Anjos committed
97 98 99


def split_four_column(filename):
100 101 102 103 104
  """split_four_column(filename) -> negatives, positives

  Loads a score set from a single file and splits the scores
  between negatives and positives. The score file has to respect the 4 column
  format as defined in the method :py:func:`four_column`.
André Anjos's avatar
André Anjos committed
105 106 107 108

  This method avoids loading and allocating memory for the strings present in
  the file. We only keep the scores.

109 110 111 112
  **Parameters:**

  ``filename`` : str or file-like
    The file that will be opened with :py:func:`open_file` containing the scores.
André Anjos's avatar
André Anjos committed
113

114 115 116 117 118 119 120 121
  **Returns:**

  ``negatives`` : array_like(1D, float)
    The list of ``score``'s, for which the ``claimed_id`` and the ``real_id`` differed (see :py:func:`four_column`).

  ``positives`` : array_like(1D, float)
    The list of ``score``'s, for which the ``claimed_id`` and the ``real_id`` are identical (see :py:func:`four_column`).
  """
122 123
  score_lines = load_score(filename, 4)
  return get_negatives_positives(score_lines)
André Anjos's avatar
André Anjos committed
124

Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
125
def cmc_four_column(filename):
126
  """cmc_four_column(filename) -> cmc_scores
127 128 129 130

  Loads scores to compute CMC curves from a file in four column format.
  The four column file needs to be in the same format as described in :py:func:`four_column`,
  and the ``test_label`` (column 3) has to contain the test/probe file name or a probe id.
131 132 133 134

  This function returns a list of tuples.
  For each probe file, the tuple consists of a list of negative scores and a list of positive scores.
  Usually, the list of positive scores should contain only one element, but more are allowed.
135
  The result of this function can directly be passed to, e.g., the :py:func:`bob.measure.cmc` function.
136

137

138 139 140 141 142
  **Parameters:**

  ``filename`` : str or file-like
    The file that will be opened with :py:func:`open_file` containing the scores.

Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
143

144 145
  **Returns:**

146 147 148
  ``cmc_scores`` : [(negatives, positives)]
    A list of tuples, where each tuple contains the ``negative`` and ``positive`` scores for one probe of the database.
    Both ``negatives`` and ``positives`` can be either an 1D :py:class:`numpy.ndarray` of type ``float``, or ``None``.
Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
149

150
  """
André Anjos's avatar
André Anjos committed
151 152 153
  # extract positives and negatives
  pos_dict = {}
  neg_dict = {}
154 155 156 157 158 159 160 161
  # read four column list
  for (client_id, probe_id, probe_name, score) in four_column(filename):
    # check in which dict we have to put the score
    correct_dict = pos_dict if client_id == probe_id else neg_dict

    # append score
    if probe_name in correct_dict:
      correct_dict[probe_name].append(score)
Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
162
    else:
163
      correct_dict[probe_name] = [score]
Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
164

165 166
  # convert that into the desired format
  return _convert_cmc_scores(neg_dict, pos_dict)
Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
167

André Anjos's avatar
André Anjos committed
168 169 170


def five_column(filename):
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
  """five_column(filename) -> claimed_id, model_label, real_id, test_label, score

  Loads a score set from a single file and yield its lines (to avoid loading the score file at once into memory).
  This function verifies that all fields are correctly placed and contain valid fields.
  The score file must contain the following information in each line:

    claimed_id model_label real_id test_label score

  **Parametes:**

  ``filename`` : str or file-like
    The file object that will be opened with :py:func:`open_file` containing the scores.

  **Yields:**

  ``claimed_id`` : str
    The claimed identity -- the client name of the model that was used in the comparison

  ``model_label`` : str
    A label for the model -- usually the model file name, or the model id

  ``real_id`` : str
    The real identity -- the client name of the probe that was used in the comparison

  ``test_label`` : str
    A label of the probe -- usually the probe file name, or the probe id

  ``score`` : float
    The result of the comparison of the model and the probe.
André Anjos's avatar
André Anjos committed
200 201
  """

202
  for i, l in enumerate(open_file(filename)):
203
    if isinstance(l, bytes): l = l.decode('utf-8')
André Anjos's avatar
André Anjos committed
204 205 206 207 208 209 210 211 212
    s = l.strip()
    if len(s) == 0 or s[0] == '#': continue #empty or comment
    field = [k.strip() for k in s.split()]
    if len(field) < 5:
      raise SyntaxError('Line %d of file "%s" is invalid: %s' % (i, filename, l))
    try:
      score = float(field[4])
    except:
      raise SyntaxError('Cannot convert score to float at line %d of file "%s": %s' % (i, filename, l))
213
    yield (field[0], field[1], field[2], field[3], score)
André Anjos's avatar
André Anjos committed
214

Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
215

André Anjos's avatar
André Anjos committed
216
def split_five_column(filename):
217 218 219 220 221
  """split_five_column(filename) -> negatives, positives

  Loads a score set from a single file in five column format and splits the scores
  between negatives and positives. The score file has to respect the 4 column
  format as defined in the method :py:func:`five_column`.
André Anjos's avatar
André Anjos committed
222 223 224 225

  This method avoids loading and allocating memory for the strings present in
  the file. We only keep the scores.

226 227 228 229 230 231 232 233 234 235 236 237
  **Parameters:**

  ``filename`` : str or file-like
    The file that will be opened with :py:func:`open_file` containing the scores.

  **Returns:**

  ``negatives`` : array_like(1D, float)
    The list of ``score``'s, for which the ``claimed_id`` and the ``real_id`` differed (see :py:func:`five_column`).

  ``positives`` : array_like(1D, float)
    The list of ``score``'s, for which the ``claimed_id`` and the ``real_id`` are identical (see :py:func:`five_column`).
André Anjos's avatar
André Anjos committed
238
  """
239 240
  score_lines = load_score(filename, 5)
  return get_negatives_positives(score_lines)
André Anjos's avatar
André Anjos committed
241

242

Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
243
def cmc_five_column(filename):
244
  """cmc_four_column(filename) -> cmc_scores
245

246 247 248
  Loads scores to compute CMC curves from a file in five column format.
  The four column file needs to be in the same format as described in :py:func:`five_column`,
  and the ``test_label`` (column 4) has to contain the test/probe file name or a probe id.
André Anjos's avatar
André Anjos committed
249

250 251 252
  This function returns a list of tuples.
  For each probe file, the tuple consists of a list of negative scores and a list of positive scores.
  Usually, the list of positive scores should contain only one element, but more are allowed.
253
  The result of this function can directly be passed to, e.g., the :py:func:`bob.measure.cmc` function.
254 255 256 257 258 259 260 261 262 263

  **Parameters:**

  ``filename`` : str or file-like
    The file that will be opened with :py:func:`open_file` containing the scores.

  **Returns:**

  ``cmc_scores`` : [(array_like(1D, float), array_like(1D, float))]
    A list of tuples, where each tuple contains the ``negative`` and ``positive`` scores for one probe of the database
Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
264

André Anjos's avatar
André Anjos committed
265
  """
266
  # extract positives and negatives
André Anjos's avatar
André Anjos committed
267 268
  pos_dict = {}
  neg_dict = {}
269 270 271
  # read four column list
  for (client_id, _, probe_id, probe_name, score) in five_column(filename):
    # check in which dict we have to put the score
272 273
    correct_dict = pos_dict if client_id == probe_id else neg_dict

274 275 276 277 278
    # append score
    if probe_name in correct_dict:
      correct_dict[probe_name].append(score)
    else:
      correct_dict[probe_name] = [score]
André Anjos's avatar
André Anjos committed
279

280 281
  # convert that into the desired format
  return _convert_cmc_scores(neg_dict, pos_dict)
Tiago de Freitas Pereira's avatar
Tiago de Freitas Pereira committed
282

283

284
def load_score(filename, ncolumns=None):
285 286 287 288 289 290 291 292
  """Load scores using numpy.loadtxt and return the data as a numpy array.

  **Parameters:**

  ``filename`` : str or file-like
    A path or file-like object that will be read with :py:func:`numpy.loadtxt`
    containing the scores.

293 294 295
  ``ncolumns`` : 4 or 5 or None [default: None]
    Specify the number of columns in the score file. If None is provided,
     the number of columns will be guessed.
296 297 298

  **Returns:**

299
  ``score_lines`` : numpy.array
300 301 302 303 304
    An array which contains not only the actual scores but also the
    'claimed_id', 'real_id', 'test_label', and ['model_label']

  """

305 306 307 308 309 310 311 312 313 314 315 316 317 318
  def convertfunc(x):
    return x

  if ncolumns not in (4, 5):
    f = open_file(filename)
    try:
      line = f.readline()
      ncolumns = len(line.split())
    except Exception:
      logger.warn('Could not guess the number of columns in file: {}. '
                  'Assuming 4 column format.'.format(filename))
      ncolumns = 4
    finally:
      f.close()
319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390

  if ncolumns == 4:
    names = ('claimed_id', 'real_id', 'test_label', 'score')
    converters = {
      0: convertfunc,
      1: convertfunc,
      2: convertfunc,
      3: float}

  elif ncolumns == 5:
    names = ('claimed_id', 'model_label', 'real_id', 'test_label', 'score')
    converters = {
      0: convertfunc,
      1: convertfunc,
      2: convertfunc,
      3: convertfunc,
      4: float}
  else:
    raise ValueError("ncolumns of 4 and 5 are supported only.")

  score_lines = numpy.genfromtxt(
    open_file(filename, mode='rb'), dtype=None, names=names,
    converters=converters, invalid_raise=True)
  new_dtype = []
  for name in score_lines.dtype.names[:-1]:
    new_dtype.append((name, str(score_lines.dtype[name]).replace('S', 'U')))
  new_dtype.append(('score', float))
  score_lines = numpy.array(score_lines, new_dtype)
  return score_lines


def get_negatives_positives(score_lines):
  """Take the output of load_score and return negatives and positives.
  This function aims to replace split_four_column and split_five_column
  but takes a different input. It's up to you to use which one.
  """
  pos_mask = score_lines['claimed_id'] == score_lines['real_id']
  positives = score_lines['score'][pos_mask]
  negatives = score_lines['score'][numpy.logical_not(pos_mask)]
  return (negatives, positives)


def get_negatives_positives_all(score_lines_list):
  """Take a list of outputs of load_score and return stacked negatives and
  positives."""
  negatives, positives = [], []
  for score_lines in score_lines_list:
    neg_pos = get_negatives_positives(score_lines)
    negatives.append(neg_pos[0])
    positives.append(neg_pos[1])
  negatives = numpy.vstack(negatives).T
  positives = numpy.vstack(positives).T
  return (negatives, positives)


def get_all_scores(score_lines_list):
  """Take a list of outputs of load_score and return stacked scores"""
  return numpy.vstack([score_lines['score']
                       for score_lines in score_lines_list]).T


def dump_score(filename, score_lines):
  """Dump scores that were loaded using :py:func:`load_score`
  The number of columns is automatically detected.
  """
  if len(score_lines.dtype) == 5:
    fmt = '%s %s %s %s %.9f'
  elif len(score_lines.dtype) == 4:
    fmt = '%s %s %s %.9f'
  else:
    raise ValueError("Only scores with 4 and 5 columns are supported.")
  numpy.savetxt(filename, score_lines, fmt=fmt)
391 392 393 394 395 396 397 398 399 400

def _convert_cmc_scores(neg_dict, pos_dict):
  """Converts the negative and positive scores read with :py:func:`cmc_four_column` or :py:func:`cmc_four_column` into a format that is handled by the :py:func:`bob.measure.cmc` and similar functions."""
  # convert to lists of tuples of ndarrays (or None)
  probe_names = sorted(set(neg_dict.keys()).union(set(pos_dict.keys())))
  # get all scores in the desired format
  return [(
    numpy.array(neg_dict[probe_name], numpy.float64) if probe_name in neg_dict else None,
    numpy.array(pos_dict[probe_name], numpy.float64) if probe_name in pos_dict else None
  ) for probe_name in probe_names]