def results_to_list(validation_results):
return [
precision(result.true_positives, result.false_positives)
for result in validation_results
]
def results_to_list(validation_results):
return [precision(result.true_positives,result.false_positives) for result in validation_results]
Xs = apply_gaussian(Xtest, sigma)
# Run each validator with the given data:
experiment_description = "%s (iteration=%s, sigma=%.2f)" % (
EXPERIMENT_NAME, iteration, sigma)
cv0.validate(Xtrain, ytrain, Xs, ytest, experiment_description)
# Get overall results:
true_positives = sum([
validation_result.true_positives
for validation_result in cv0.validation_results
])
false_positives = sum([
validation_result.false_positives
for validation_result in cv0.validation_results
])
# Calculate overall precision:
prec = precision(true_positives, false_positives)
# Store the result:
print(key)
experiments[key]['results'][sigma] = prec
# Make a nice plot of this textual output:
fig = plt.figure()
# Holds the legend items:
plot_legend = []
# Add the Validation results:
for experiment_name, experiment_definition in experiments.iteritems():
print(key, experiment_definition)
results = experiment_definition['results']
(xvalues, yvalues) = zip(*[(k, v) for k, v in results.iteritems()])
# Add to the legend:
plot_legend.append(experiment_name)
# Define the validators for the model:
cv0 = SimpleValidation(value['model'])
for iteration in xrange(ITER_MAX):
print "Repeating experiment %s/%s." % (iteration + 1, ITER_MAX)
# Split dataset according to the papers description:
Xtrain, ytrain, Xtest, ytest = partition_data(X,y)
# Apply a gaussian blur on the images:
Xs = apply_gaussian(Xtest, sigma)
# Run each validator with the given data:
experiment_description = "%s (iteration=%s, sigma=%.2f)" % (EXPERIMENT_NAME, iteration, sigma)
cv0.validate(Xtrain, ytrain, Xs, ytest, experiment_description)
# Get overall results:
true_positives = sum([validation_result.true_positives for validation_result in cv0.validation_results])
false_positives = sum([validation_result.false_positives for validation_result in cv0.validation_results])
# Calculate overall precision:
prec = precision(true_positives,false_positives)
# Store the result:
print key
experiments[key]['results'][sigma] = prec
# Make a nice plot of this textual output:
fig = plt.figure()
# Holds the legend items:
plot_legend = []
# Add the Validation results:
for experiment_name, experiment_definition in experiments.iteritems():
print key, experiment_definition
results = experiment_definition['results']
(xvalues, yvalues) = zip(*[(k,v) for k,v in results.iteritems()])
# Add to the legend:
plot_legend.append(experiment_name)
