Merge branch 'multiple-protocols' into 'master'

Add a command for multi protocol (N-fold cross validation) analysis

See merge request !79

bob/measure/script/commands.py

@@ -11,10 +11,12 @@ SCORE_FORMAT = (
 CRITERIA = ('eer', 'min-hter', 'far')
 
 
-@common_options.metrics_command(common_options.METRICS_HELP.format(
-    names='FtA, FAR, FRR, FMR, FMNR, HTER',
-    criteria=CRITERIA, score_format=SCORE_FORMAT,
-    command='bob measure metrics'), criteria=CRITERIA)
+@common_options.metrics_command(
+    common_options.METRICS_HELP.format(
+        names='FtA, FAR, FRR, FMR, FMNR, HTER',
+        criteria=CRITERIA, score_format=SCORE_FORMAT,
+        command='bob measure metrics'),
+    criteria=CRITERIA)
 def metrics(ctx, scores, evaluation, **kwargs):
     process = figure.Metrics(ctx, scores, evaluation, load.split)
     process.run()
@@ -59,3 +61,15 @@ def hist(ctx, scores, evaluation, **kwargs):
 def evaluate(ctx, scores, evaluation, **kwargs):
     common_options.evaluate_flow(
         ctx, scores, evaluation, metrics, roc, det, epc, hist, **kwargs)
+
+
+@common_options.multi_metrics_command(
+    common_options.MULTI_METRICS_HELP.format(
+        names='FtA, FAR, FRR, FMR, FMNR, HTER',
+        criteria=CRITERIA, score_format=SCORE_FORMAT,
+        command='bob measure multi-metrics'),
+    criteria=CRITERIA)
+def multi_metrics(ctx, scores, evaluation, protocols_number, **kwargs):
+    ctx.meta['min_arg'] = protocols_number * (2 if evaluation else 1)
+    process = figure.MultiMetrics(ctx, scores, evaluation, load.split)
+    process.run()

bob/measure/script/common_options.py

@@ -21,7 +21,7 @@ def scores_argument(min_arg=1, force_eval=False, **kwargs):
     ----------
     min_arg : int
         the minimum number of file needed to evaluate a system. For example,
-        PAD functionalities needs licit abd spoof and therefore min_arg = 2
+        vulnerability analysis needs licit and spoof and therefore min_arg = 2
 
     Returns
     -------
@@ -920,3 +920,67 @@ def evaluate_flow(ctx, scores, evaluation, metrics, roc, det, epc, hist,
     ctx.forward(hist)
     click.echo("Evaluate successfully completed!")
     click.echo("[plots] => %s" % (ctx.meta['output']))
+
+
+def n_protocols_option(required=True, **kwargs):
+    '''Get option for number of protocols.'''
+    def custom_n_protocols_option(func):
+        def callback(ctx, param, value):
+            value = abs(value)
+            ctx.meta['protocols_number'] = value
+            return value
+        return click.option(
+            '-pn', '--protocols-number', type=click.INT,
+            show_default=True, required=required,
+            help='The number of protocols of cross validation.',
+            callback=callback, **kwargs)(func)
+    return custom_n_protocols_option
+
+
+def multi_metrics_command(docstring, criteria=('eer', 'min-hter', 'far')):
+    def custom_metrics_command(func):
+        func.__doc__ = docstring
+
+        @click.command('multi-metrics')
+        @scores_argument(nargs=-1)
+        @eval_option()
+        @n_protocols_option()
+        @table_option()
+        @output_log_metric_option()
+        @criterion_option(criteria)
+        @thresholds_option()
+        @far_option()
+        @legends_option()
+        @open_file_mode_option()
+        @verbosity_option()
+        @click.pass_context
+        @functools.wraps(func)
+        def wrapper(*args, **kwds):
+            return func(*args, **kwds)
+        return wrapper
+    return custom_metrics_command
+
+
+MULTI_METRICS_HELP = """Multi protocol (cross-validation) metrics.
+
+    Prints a table that contains {names} for a given threshold criterion
+    ({criteria}). The metrics are averaged over several protocols. The idea is
+    that each protocol corresponds to one fold in your cross-validation.
+
+    You need to provide as many as development score files as the number of
+    protocols per system. You can also provide evaluation files along with dev
+    files. If evaluation scores are provided, you must use flag `--eval`. The
+    number of protocols must be provided using the `--protocols-number` option.
+
+    {score_format}
+
+    Resulting table format can be changed using the `--tablefmt`.
+
+    Examples:
+
+        $ {command} -v {{p1,p2,p3}}/scores-dev
+
+        $ {command} -v -e {{p1,p2,p3}}/scores-{{dev,eval}}
+
+        $ {command} -v -e {{sys1,sys2}}/{{p1,p2,p3}}/scores-{{dev,eval}}
+    """

bob/measure/script/figure.py

@@ -5,6 +5,7 @@ from abc import ABCMeta, abstractmethod
 import math
 import sys
 import os.path
+import numpy
 import click
 import matplotlib
 import matplotlib.pyplot as mpl
@@ -128,7 +129,6 @@ class MeasureBase(object):
         # and if only dev:
         # [ (dev_licit_neg, dev_licit_pos), (dev_spoof_neg, dev_licit_pos)]
 
-
     # Things to do after the main iterative computations are done
     @abstractmethod
     def end_process(self):
@@ -192,6 +192,29 @@ class Metrics(MeasureBase):
     def get_thres(self, criterion, dev_neg, dev_pos, far):
         return utils.get_thres(criterion, dev_neg, dev_pos, far)
 
+    def _numbers(self, neg, pos, threshold, fta):
+        from .. import farfrr
+        fmr, fnmr = farfrr(neg, pos, threshold)
+        far = fmr * (1 - fta)
+        frr = fta + fnmr * (1 - fta)
+        hter = (far + frr) / 2.0
+
+        ni = neg.shape[0]  # number of impostors
+        fm = int(round(fmr * ni))  # number of false accepts
+        nc = pos.shape[0]  # number of clients
+        fnm = int(round(fnmr * nc))  # number of false rejects
+        return fta, fmr, fnmr, hter, far, frr, fm, ni, fnm, nc
+
+    def _strings(self, fta, fmr, fnmr, hter, far, frr, fm, ni, fnm, nc):
+        fta_str = "%.1f%%" % (100 * fta)
+        fmr_str = "%.1f%% (%d/%d)" % (100 * fmr, fm, ni)
+        fnmr_str = "%.1f%% (%d/%d)" % (100 * fnmr, fnm, nc)
+        far_str = "%.1f%%" % (100 * far)
+        frr_str = "%.1f%%" % (100 * frr)
+        hter_str = "%.1f%%" % (100 * hter)
+
+        return fta_str, fmr_str, fnmr_str, far_str, frr_str, hter_str
+
     def compute(self, idx, input_scores, input_names):
         ''' Compute metrics thresholds and tables (FAR, FMR, FNMR, HTER) for
         given system inputs'''
@@ -204,6 +227,7 @@ class Metrics(MeasureBase):
 
         threshold = self.get_thres(self._criterion, dev_neg, dev_pos, self._far) \
             if self._thres is None else self._thres[idx]
+
         title = self._legends[idx] if self._legends is not None else None
         if self._thres is None:
             far_str = ''
@@ -219,63 +243,33 @@ class Metrics(MeasureBase):
                        "Development set `%s`: %e"
                        % (dev_file or title, threshold), file=self.log_file)
 
-        from .. import farfrr
-        dev_fmr, dev_fnmr = farfrr(dev_neg, dev_pos, threshold)
-        dev_far = dev_fmr * (1 - dev_fta)
-        dev_frr = dev_fta + dev_fnmr * (1 - dev_fta)
-        dev_hter = (dev_far + dev_frr) / 2.0
-
-        dev_ni = dev_neg.shape[0]  # number of impostors
-        dev_fm = int(round(dev_fmr * dev_ni))  # number of false accepts
-        dev_nc = dev_pos.shape[0]  # number of clients
-        dev_fnm = int(round(dev_fnmr * dev_nc))  # number of false rejects
-
-        dev_fta_str = "%.1f%%" % (100 * dev_fta)
-        dev_fmr_str = "%.1f%% (%d/%d)" % (100 * dev_fmr, dev_fm, dev_ni)
-        dev_fnmr_str = "%.1f%% (%d/%d)" % (100 * dev_fnmr, dev_fnm, dev_nc)
-        dev_far_str = "%.1f%%" % (100 * dev_far)
-        dev_frr_str = "%.1f%%" % (100 * dev_frr)
-        dev_hter_str = "%.1f%%" % (100 * dev_hter)
+        fta_str, fmr_str, fnmr_str, far_str, frr_str, hter_str = \
+            self._strings(*self._numbers(
+                dev_neg, dev_pos, threshold, dev_fta))
         headers = ['' or title, 'Development %s' % dev_file]
-        raws = [[self.names[0], dev_fta_str],
-                [self.names[1], dev_fmr_str],
-                [self.names[2], dev_fnmr_str],
-                [self.names[3], dev_far_str],
-                [self.names[4], dev_frr_str],
-                [self.names[5], dev_hter_str]]
+        rows = [[self.names[0], fta_str],
+                [self.names[1], fmr_str],
+                [self.names[2], fnmr_str],
+                [self.names[3], far_str],
+                [self.names[4], frr_str],
+                [self.names[5], hter_str]]
 
         if self._eval:
-            # computes statistics for the eval set based on the threshold a priori
-            eval_fmr, eval_fnmr = farfrr(eval_neg, eval_pos, threshold)
-            eval_far = eval_fmr * (1 - eval_fta)
-            eval_frr = eval_fta + eval_fnmr * (1 - eval_fta)
-            eval_hter = (eval_far + eval_frr) / 2.0
-
-            eval_ni = eval_neg.shape[0]  # number of impostors
-            eval_fm = int(round(eval_fmr * eval_ni))  # number of false accepts
-            eval_nc = eval_pos.shape[0]  # number of clients
-            # number of false rejects
-            eval_fnm = int(round(eval_fnmr * eval_nc))
-
-            eval_fta_str = "%.1f%%" % (100 * eval_fta)
-            eval_fmr_str = "%.1f%% (%d/%d)" % (100 *
-                                               eval_fmr, eval_fm, eval_ni)
-            eval_fnmr_str = "%.1f%% (%d/%d)" % (100 *
-                                                eval_fnmr, eval_fnm, eval_nc)
-
-            eval_far_str = "%.1f%%" % (100 * eval_far)
-            eval_frr_str = "%.1f%%" % (100 * eval_frr)
-            eval_hter_str = "%.1f%%" % (100 * eval_hter)
+            # computes statistics for the eval set based on the threshold a
+            # priori
+            fta_str, fmr_str, fnmr_str, far_str, frr_str, hter_str = \
+                self._strings(*self._numbers(
+                    eval_neg, eval_pos, threshold, eval_fta))
 
             headers.append('Eval. % s' % eval_file)
-            raws[0].append(eval_fta_str)
-            raws[1].append(eval_fmr_str)
-            raws[2].append(eval_fnmr_str)
-            raws[3].append(eval_far_str)
-            raws[4].append(eval_frr_str)
-            raws[5].append(eval_hter_str)
+            rows[0].append(fta_str)
+            rows[1].append(fmr_str)
+            rows[2].append(fnmr_str)
+            rows[3].append(far_str)
+            rows[4].append(frr_str)
+            rows[5].append(hter_str)
 
-        click.echo(tabulate(raws, headers, self._tablefmt), file=self.log_file)
+        click.echo(tabulate(rows, headers, self._tablefmt), file=self.log_file)
 
     def end_process(self):
         ''' Close log file if needed'''
@@ -283,6 +277,91 @@ class Metrics(MeasureBase):
             self.log_file.close()
 
 
+class MultiMetrics(Metrics):
+    '''Computes average of metrics based on several protocols (cross
+    validation)
+
+    Attributes
+    ----------
+    log_file : str
+        output stream
+    names : tuple
+        List of names for the metrics.
+    '''
+
+    def __init__(self, ctx, scores, evaluation, func_load,
+                 names=('NaNs Rate', 'False Positive Rate',
+                        'False Negative Rate', 'False Accept Rate',
+                        'False Reject Rate', 'Half Total Error Rate')):
+        super(MultiMetrics, self).__init__(
+            ctx, scores, evaluation, func_load, names=names)
+
+        self.headers = ['Methods'] + list(self.names)
+        if self._eval:
+            self.headers.insert(1, self.names[5] + ' (dev)')
+        self.rows = []
+
+    def _strings(self, metrics):
+        ftam, fmrm, fnmrm, hterm, farm, frrm, _, _, _, _ = metrics.mean(axis=0)
+        ftas, fmrs, fnmrs, hters, fars, frrs, _, _, _, _ = metrics.std(axis=0)
+        fta_str = "%.1f%% (%.1f%%)" % (100 * ftam, 100 * ftas)
+        fmr_str = "%.1f%% (%.1f%%)" % (100 * fmrm, 100 * fmrs)
+        fnmr_str = "%.1f%% (%.1f%%)" % (100 * fnmrm, 100 * fnmrs)
+        far_str = "%.1f%% (%.1f%%)" % (100 * farm, 100 * fars)
+        frr_str = "%.1f%% (%.1f%%)" % (100 * frrm, 100 * frrs)
+        hter_str = "%.1f%% (%.1f%%)" % (100 * hterm, 100 * hters)
+
+        return fta_str, fmr_str, fnmr_str, far_str, frr_str, hter_str
+
+    def compute(self, idx, input_scores, input_names):
+        '''Computes the average of metrics over several protocols.'''
+        neg_list, pos_list, fta_list = utils.get_fta_list(input_scores)
+        step = 2 if self._eval else 1
+        self._dev_metrics = []
+        self._thresholds = []
+        for i in range(0, len(input_scores), step):
+            neg, pos, fta = neg_list[i], pos_list[i], fta_list[i]
+            threshold = self.get_thres(self._criterion, neg, pos, self._far) \
+                if self._thres is None else self._thres[idx]
+            self._thresholds.append(threshold)
+            self._dev_metrics.append(self._numbers(neg, pos, threshold, fta))
+        self._dev_metrics = numpy.array(self._dev_metrics)
+
+        if self._eval:
+            self._eval_metrics = []
+            for i in range(1, len(input_scores), step):
+                neg, pos, fta = neg_list[i], pos_list[i], fta_list[i]
+                threshold = self._thresholds[i // 2]
+                self._eval_metrics.append(
+                    self._numbers(neg, pos, threshold, fta))
+            self._eval_metrics = numpy.array(self._eval_metrics)
+
+        title = self._legends[idx] if self._legends is not None else None
+
+        fta_str, fmr_str, fnmr_str, far_str, frr_str, hter_str = \
+            self._strings(self._dev_metrics)
+
+        if self._eval:
+            self.rows.append([title, hter_str])
+        else:
+            self.rows.append([title, fta_str, fmr_str, fnmr_str,
+                              far_str, frr_str, hter_str])
+
+        if self._eval:
+            # computes statistics for the eval set based on the threshold a
+            # priori
+            fta_str, fmr_str, fnmr_str, far_str, frr_str, hter_str = \
+                self._strings(self._eval_metrics)
+
+            self.rows[-1].extend([fta_str, fmr_str, fnmr_str,
+                                  far_str, frr_str, hter_str])
+
+    def end_process(self):
+        click.echo(tabulate(self.rows, self.headers,
+                            self._tablefmt), file=self.log_file)
+        super(MultiMetrics, self).end_process()
+
+
 class PlotBase(MeasureBase):
     ''' Base class for plots. Regroup several options and code
     shared by the different plots
@@ -586,7 +665,8 @@ class Hist(PlotBase):
         # do not display dev histo
         self._hide_dev = ctx.meta.get('hide_dev', False)
         if self._hide_dev and not self._eval:
-            raise click.BadParameter("You can only use --hide-dev along with --eval")
+            raise click.BadParameter(
+                "You can only use --hide-dev along with --eval")
 
         # dev hist are displayed next to eval hist
         self._ncols *= 1 if self._hide_dev or not self._eval else 2
@@ -601,7 +681,7 @@ class Hist(PlotBase):
         if self._legends is not None and len(self._legends) == self.n_systems \
            and not self._hide_dev:
             # use same legend for dev and eval if needed
-            self._legends = [x for pair in zip(self._legends,self._legends)
+            self._legends = [x for pair in zip(self._legends, self._legends)
                              for x in pair]
 
     def compute(self, idx, input_scores, input_names):

setup.py

@@ -74,6 +74,7 @@ setup(
       'bob.measure.cli': [
           'evaluate = bob.measure.script.commands:evaluate',
           'metrics = bob.measure.script.commands:metrics',
+          'multi-metrics = bob.measure.script.commands:multi_metrics',
           'roc = bob.measure.script.commands:roc',
           'det = bob.measure.script.commands:det',
           'epc = bob.measure.script.commands:epc',