def partition_data(X, y):
"""
Shuffles the input data and splits it into a new set of images. This resembles the experimental setup
used in the paper on the Local Phase Quantization descriptor in:
"Recognition of Blurred Faces Using Local Phase Quantization", Timo Ahonen, Esa Rahtu, Ville Ojansivu, Janne Heikkila
What it does is to build a subset for each class, so it has 1 image for training and the rest for testing.
The original dataset is shuffled for each call, hence you always get a new partitioning.
"""
Xs,ys = shuffle_array(X,y)
# Maps index to class:
mapping = {}
for i in xrange(len(y)):
yi = ys[i]
try:
mapping[yi].append(i)
except KeyError:
mapping[yi] = [i]
# Get one image for each subject:
Xtrain, ytrain = [], []
Xtest, ytest = [], []
# Finally build partition:
for key, indices in mapping.iteritems():
# Add images:
Xtrain.extend([ Xs[i] for i in indices[:1] ])
ytrain.extend([ ys[i] for i in indices[:1] ])
Xtest.extend([ Xs[i] for i in indices[1:20]])
ytest.extend([ ys[i] for i in indices[1:20]])
# Return shuffled partitions:
return Xtrain, ytrain, Xtest, ytest
def partition_data(X, y):
"""
Shuffles the input data and splits it into a new set of images. This resembles the experimental setup
used in the paper on the Local Phase Quantization descriptor in:
"Recognition of Blurred Faces Using Local Phase Quantization", Timo Ahonen, Esa Rahtu, Ville Ojansivu, Janne Heikkila
What it does is to build a subset for each class, so it has 1 image for training and the rest for testing.
The original dataset is shuffled for each call, hence you always get a new partitioning.
"""
Xs, ys = shuffle_array(X, y)
# Maps index to class:
mapping = {}
for i in xrange(len(y)):
yi = ys[i]
try:
mapping[yi].append(i)
except KeyError:
mapping[yi] = [i]
# Get one image for each subject:
Xtrain, ytrain = [], []
Xtest, ytest = [], []
# Finally build partition:
for key, indices in mapping.iteritems():
# Add images:
Xtrain.extend([Xs[i] for i in indices[:1]])
ytrain.extend([ys[i] for i in indices[:1]])
Xtest.extend([Xs[i] for i in indices[1:20]])
ytest.extend([ys[i] for i in indices[1:20]])
# Return shuffled partitions:
return Xtrain, ytrain, Xtest, ytest
def partition_data(X, y):
Xs,ys = shuffle_array(X,y)
mapping = {}
for i in xrange(len(y)):
yi = ys[i]
try:
mapping[yi].append(i)
except KeyError:
mapping[yi] = [i]
Xtrain, ytrain = [], []
Xtest, ytest = [], []
for key, indices in mapping.iteritems():
Xtrain.extend([ Xs[i] for i in indices[:1] ])
ytrain.extend([ ys[i] for i in indices[:1] ])
Xtest.extend([ Xs[i] for i in indices[1:20]])
ytest.extend([ ys[i] for i in indices[1:20]])
return Xtrain, ytrain, Xtest, ytest