def test_unit_norm(self):
""" Test that using std_bias = 0.0 and use_norm = True
results in vectors having unit norm """
tol = 1e-5
num_examples = 5
num_features = 10
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(num_examples, num_features))
dataset = DenseDesignMatrix(X=X)
# the setting of subtract_mean is not relevant to the test
# the test only applies when std_bias = 0.0 and use_std = False
preprocessor = GlobalContrastNormalization(subtract_mean=False,
sqrt_bias=0.0,
use_std=False)
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
norms = np.sqrt(np.square(result).sum(axis=1))
max_norm_error = np.abs(norms - 1.).max()
tol = 3e-5
assert max_norm_error < tol
def test_extract_reassemble():
""" Tests that ExtractGridPatches and ReassembleGridPatches are
inverse of each other """
rng = np.random.RandomState([1, 3, 7])
topo = rng.randn(4, 3 * 5, 3 * 7, 2)
dataset = DenseDesignMatrix(topo_view=topo)
patch_shape = (3, 7)
extractor = ExtractGridPatches(patch_shape, patch_shape)
reassemblor = ReassembleGridPatches(patch_shape=patch_shape,
orig_shape=topo.shape[1:3])
dataset.apply_preprocessor(extractor)
dataset.apply_preprocessor(reassemblor)
new_topo = dataset.get_topological_view()
assert new_topo.shape == topo.shape
if not np.all(new_topo == topo):
assert False
def test_extract_reassemble():
""" Tests that ExtractGridPatches and ReassembleGridPatches are
inverse of each other """
rng = np.random.RandomState([1, 3, 7])
topo = rng.randn(4, 3 * 5, 3 * 7, 2)
dataset = DenseDesignMatrix(topo_view=topo)
patch_shape = (3, 7)
extractor = ExtractGridPatches(patch_shape, patch_shape)
reassemblor = ReassembleGridPatches(patch_shape=patch_shape,
orig_shape=topo.shape[1:3])
dataset.apply_preprocessor(extractor)
dataset.apply_preprocessor(reassemblor)
new_topo = dataset.get_topological_view()
assert new_topo.shape == topo.shape
if not np.all(new_topo == topo):
assert False
def test_unit_norm(self):
""" Test that using std_bias = 0.0 and use_norm = True
results in vectors having unit norm """
tol = 1e-5
num_examples = 5
num_features = 10
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(num_examples, num_features))
dataset = DenseDesignMatrix(X=X)
# the setting of subtract_mean is not relevant to the test
# the test only applies when std_bias = 0.0 and use_std = False
preprocessor = GlobalContrastNormalization(subtract_mean=False,
sqrt_bias=0.0,
use_std=False)
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
norms = np.sqrt(np.square(result).sum(axis=1))
max_norm_error = np.abs(norms - 1.).max()
tol = 3e-5
assert max_norm_error < tol
def test_zero_image(self):
"""
Test on zero-value image if cause any division by zero
"""
X = as_floatX(np.zeros((5, 32 * 32 * 3)))
axes = ['b', 0, 1, 'c']
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3),
axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = LeCunLCN(img_shape=[32, 32])
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert isfinite(result)
def test_zero_image(self):
"""
Test on zero-value image if cause any division by zero
"""
X = as_floatX(np.zeros((5, 32 * 32 * 3)))
axes = ['b', 0, 1, 'c']
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3),
axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = LeCunLCN(img_shape=[32, 32])
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert isfinite(result)
def test_channel(self):
"""
Test if works fine withe different number of channel as argument
"""
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(5, 32 * 32 * 3))
axes = ['b', 0, 1, 'c']
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3),
axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = LeCunLCN(img_shape=[32, 32], channels=[1, 2])
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert isfinite(result)
def test_channel(self):
"""
Test if works fine withe different number of channel as argument
"""
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(5, 32 * 32 * 3))
axes = ['b', 0, 1, 'c']
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3),
axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = LeCunLCN(img_shape=[32, 32], channels=[1, 2])
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert isfinite(result)
def test_zero_vector(self):
""" Test that passing in the zero vector does not result in
a divide by 0 """
dataset = DenseDesignMatrix(X=as_floatX(np.zeros((1, 1))))
# the settings of subtract_mean and use_norm are not relevant to
# the test
# std_bias = 0.0 is the only value for which there should be a risk
# of failure occurring
preprocessor = GlobalContrastNormalization(subtract_mean=True, sqrt_bias=0.0, use_std=True)
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert not np.any(np.isnan(result))
assert not np.any(np.isinf(result))
def test_zero_vector(self):
""" Test that passing in the zero vector does not result in
a divide by 0 """
dataset = DenseDesignMatrix(X=as_floatX(np.zeros((1, 1))))
# the settings of subtract_mean and use_norm are not relevant to
# the test
# std_bias = 0.0 is the only value for which there should be a risk
# of failure occurring
preprocessor = GlobalContrastNormalization(subtract_mean=True,
sqrt_bias=0.0,
use_std=True)
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert isfinite(result)
def test_rgb_yuv():
"""
Test on a random image if the per-processor loads and works without
anyerror and doesn't result in any nan or inf values
"""
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(5, 32 * 32 * 3))
axes = ['b', 0, 1, 'c']
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3),
axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = RGB_YUV()
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert isfinite(result)
def test_rgb_yuv():
"""
Test on a random image if the per-processor loads and works without
anyerror and doesn't result in any nan or inf values
"""
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(5, 32 * 32 * 3))
axes = ['b', 0, 1, 'c']
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3),
axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = RGB_YUV()
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert isfinite(result)
def test_random_image(self):
"""
Test on a random image if the per-processor loads and works without
anyerror and doesn't result in any nan or inf values
"""
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(5, 32 * 32 * 3))
axes = ["b", 0, 1, "c"]
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3), axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = LeCunLCN(img_shape=[32, 32])
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert not np.any(np.isnan(result))
assert not np.any(np.isinf(result))
def test_random_image(self):
"""
Test on a random image if the per-processor loads and works without
anyerror and doesn't result in any nan or inf values
"""
rng = np.random.RandomState([1, 2, 3])
X = as_floatX(rng.randn(5, 32 * 32 * 3))
axes = ['b', 0, 1, 'c']
view_converter = dense_design_matrix.DefaultViewConverter((32, 32, 3),
axes)
dataset = DenseDesignMatrix(X=X, view_converter=view_converter)
dataset.axes = axes
preprocessor = LeCunLCN(img_shape=[32, 32])
dataset.apply_preprocessor(preprocessor)
result = dataset.get_design_matrix()
assert not np.any(np.isnan(result))
assert not np.any(np.isinf(result))
from pylearn2.utils import serial
stl10 = serial.load('/data/lisa/data/stl10/stl10_32x32/train.pkl')
batch = stl10.X[24:25,:]
img = stl10.view_converter.design_mat_to_topo_view(batch)[0,...] / 127.5
from pylearn2.gui.patch_viewer import PatchViewer
pv = PatchViewer((27,27),(6,6),pad=(1,1),is_color=True)
pipeline = serial.load('/data/lisa/data/stl10/stl10_patches/preprocessor.pkl')
del pipeline.items[0]
from pylearn2.datasets.dense_design_matrix import DenseDesignMatrix, DefaultViewConverter
for row in xrange(27):
for col in xrange(27):
patch = img[row:row+6,col:col+6]
d = DenseDesignMatrix( topo_view = patch.reshape(1,6,6,3), view_converter = DefaultViewConverter((6,6,3)) )
d.apply_preprocessor(pipeline)
pv.add_patch(d.get_topological_view()[0,...], rescale = True)
pv.show()
rng = np.random.RandomState([1,2,3])
for i, img_path in enumerate(ImageIterator(path, suffix=".npy")):
img = np.load(img_path)
if img.shape[2] == 1:
img = np.concatenate((img,img,img),axis=2)
img = img.reshape(1,img.shape[0],img.shape[1],img.shape[2])
d = DenseDesignMatrix( topo_view = img, view_converter = DefaultViewConverter(img.shape[1:]) )
random_rng = np.random.RandomState([ rng.randint(0,256), rng.randint(0,256), rng.randint(0,256)])
p = ExtractPatches( patch_shape = patch_shape, num_patches = k , rng = random_rng)
d.apply_preprocessor(p)
X[i*3:(i+1)*3,:] = d.X
d.X = X
base = '/data/lisatmp/goodfeli/darpa_imagenet_patch_%dx%d_train.' % (patch_shape[0], patch_shape[1])
d.use_design_loc(base+'npy')
serial.save(base+'pkl',d)
def __call__(self, full_X):
feature_type = self.feature_type
pooling_region_counts = self.pooling_region_counts
model = self.model
size = self.size
nan = 0
full_X = full_X.reshape(1,full_X.shape[0],full_X.shape[1],full_X.shape[2])
if full_X.shape[3] == 1:
full_X = np.concatenate( (full_X, full_X, full_X), axis=3)
print 'full_X.shape: '+str(full_X.shape)
num_examples = full_X.shape[0]
assert num_examples == 1
pipeline = self.preprocessor
def average_pool( stride ):
def point( p ):
return p * ns / stride
rval = np.zeros( (topo_feat.shape[0], stride, stride, topo_feat.shape[3] ) , dtype = 'float32')
for i in xrange(stride):
for j in xrange(stride):
rval[:,i,j,:] = self.region_features( topo_feat[:,point(i):point(i+1), point(j):point(j+1),:] )
return rval
outputs = [ np.zeros((num_examples,count,count,model.nhid),dtype='float32') for count in pooling_region_counts ]
assert len(outputs) > 0
fd = DenseDesignMatrix(X = np.zeros((1,1),dtype='float32'), view_converter = DefaultViewConverter([1, 1, model.nhid] ) )
ns = 32 - size + 1
depatchifier = ReassembleGridPatches( orig_shape = (ns, ns), patch_shape=(1,1) )
batch_size = 1
for i in xrange(0,num_examples-batch_size+1,batch_size):
print i
t1 = time.time()
d = DenseDesignMatrix( topo_view = np.cast['float32'](full_X[i:i+batch_size,:]), view_converter = DefaultViewConverter((32,32,3)))
t2 = time.time()
#print '\tapplying preprocessor'
d.apply_preprocessor(pipeline, can_fit = False)
X2 = d.get_design_matrix()
t3 = time.time()
#print '\trunning theano function'
feat = self.f(X2)
t4 = time.time()
assert feat.dtype == 'float32'
feat_dataset = copy.copy(fd)
if np.any(np.isnan(feat)):
nan += np.isnan(feat).sum()
feat[np.isnan(feat)] = 0
feat_dataset.set_design_matrix(feat)
#print '\treassembling features'
feat_dataset.apply_preprocessor(depatchifier)
#print '\tmaking topological view'
topo_feat = feat_dataset.get_topological_view()
assert topo_feat.shape[0] == batch_size
t5 = time.time()
#average pooling
for output, count in zip(outputs, pooling_region_counts):
output[i:i+batch_size,...] = average_pool(count)
t6 = time.time()
print (t6-t1, t2-t1, t3-t2, t4-t3, t5-t4, t6-t5)
return outputs[0]
def __call__(self, full_X):
feature_type = self.feature_type
pooling_region_counts = self.pooling_region_counts
model = self.model
size = self.size
nan = 0
full_X = full_X.reshape(1, full_X.shape[0], full_X.shape[1],
full_X.shape[2])
if full_X.shape[3] == 1:
full_X = np.concatenate((full_X, full_X, full_X), axis=3)
print 'full_X.shape: ' + str(full_X.shape)
num_examples = full_X.shape[0]
assert num_examples == 1
pipeline = self.preprocessor
def average_pool(stride):
def point(p):
return p * ns / stride
rval = np.zeros(
(topo_feat.shape[0], stride, stride, topo_feat.shape[3]),
dtype='float32')
for i in xrange(stride):
for j in xrange(stride):
rval[:, i, j, :] = self.region_features(
topo_feat[:,
point(i):point(i + 1),
point(j):point(j + 1), :])
return rval
outputs = [
np.zeros((num_examples, count, count, model.nhid), dtype='float32')
for count in pooling_region_counts
]
assert len(outputs) > 0
fd = DenseDesignMatrix(X=np.zeros((1, 1), dtype='float32'),
view_converter=DefaultViewConverter(
[1, 1, model.nhid]))
ns = 32 - size + 1
depatchifier = ReassembleGridPatches(orig_shape=(ns, ns),
patch_shape=(1, 1))
batch_size = 1
for i in xrange(0, num_examples - batch_size + 1, batch_size):
print i
t1 = time.time()
d = DenseDesignMatrix(
topo_view=np.cast['float32'](full_X[i:i + batch_size, :]),
view_converter=DefaultViewConverter((32, 32, 3)))
t2 = time.time()
#print '\tapplying preprocessor'
d.apply_preprocessor(pipeline, can_fit=False)
X2 = d.get_design_matrix()
t3 = time.time()
#print '\trunning theano function'
feat = self.f(X2)
t4 = time.time()
assert feat.dtype == 'float32'
feat_dataset = copy.copy(fd)
if np.any(np.isnan(feat)):
nan += np.isnan(feat).sum()
feat[np.isnan(feat)] = 0
feat_dataset.set_design_matrix(feat)
#print '\treassembling features'
feat_dataset.apply_preprocessor(depatchifier)
#print '\tmaking topological view'
topo_feat = feat_dataset.get_topological_view()
assert topo_feat.shape[0] == batch_size
t5 = time.time()
#average pooling
for output, count in zip(outputs, pooling_region_counts):
output[i:i + batch_size, ...] = average_pool(count)
t6 = time.time()
print(t6 - t1, t2 - t1, t3 - t2, t4 - t3, t5 - t4, t6 - t5)
return outputs[0]
feat = ( H > 0.5) * Mu1
else:
assert False
print 'compiling theano function'
f = function([V],feat)
print 'running theano function'
feat = f(X2)
feat_dataset = DenseDesignMatrix(X = feat, view_converter = DefaultViewConverter([1, 1, feat.shape[1]] ) )
print 'reassembling features'
ns = 32 - size + 1
depatchifier = ReassembleGridPatches( orig_shape = (ns, ns), patch_shape=(1,1) )
feat_dataset.apply_preprocessor(depatchifier)
print 'making topological view'
topo_feat = feat_dataset.get_topological_view()
assert topo_feat.shape[0] == X.shape[0]
print 'assembling visualizer'
n = np.ceil(np.sqrt(model.nhid))
pv3 = PatchViewer(grid_shape = (X.shape[0], num_filters), patch_shape=(ns,ns), is_color= False)
pv4 = PatchViewer(grid_shape = (n,n), patch_shape = (size,size), is_color = True, pad = (7,7))
pv5 = PatchViewer(grid_shape = (1,num_filters), patch_shape = (size,size), is_color = True, pad = (7,7))
idx = sorted(range(model.nhid), key = lambda l : -topo_feat[:,:,:,l].std() )
print 'compiling theano function'
f = function([V], feat)
print 'running theano function'
feat = f(X2)
feat_dataset = DenseDesignMatrix(X=feat,
view_converter=DefaultViewConverter(
[1, 1, feat.shape[1]]))
print 'reassembling features'
ns = 32 - size + 1
depatchifier = ReassembleGridPatches(orig_shape=(ns, ns),
patch_shape=(1, 1))
feat_dataset.apply_preprocessor(depatchifier)
print 'making topological view'
topo_feat = feat_dataset.get_topological_view()
assert topo_feat.shape[0] == X.shape[0]
print 'assembling visualizer'
n = np.ceil(np.sqrt(model.nhid))
pv3 = PatchViewer(grid_shape=(X.shape[0], num_filters),
patch_shape=(ns, ns),
is_color=False)
pv4 = PatchViewer(grid_shape=(n, n),
patch_shape=(size, size),
is_color=True,
for i, img_path in enumerate(ImageIterator(path, suffix=".npy")):
img = np.load(img_path)
if img.shape[2] == 1:
img = np.concatenate((img, img, img), axis=2)
img = img.reshape(1, img.shape[0], img.shape[1], img.shape[2])
d = DenseDesignMatrix(topo_view=img,
view_converter=DefaultViewConverter(img.shape[1:]))
random_rng = np.random.RandomState(
[rng.randint(0, 256),
rng.randint(0, 256),
rng.randint(0, 256)])
p = ExtractPatches(patch_shape=patch_shape, num_patches=k, rng=random_rng)
d.apply_preprocessor(p)
X[i * 3:(i + 1) * 3, :] = d.X
d.X = X
base = '/data/lisatmp/goodfeli/darpa_imagenet_patch_%dx%d_train.' % (
patch_shape[0], patch_shape[1])
d.use_design_loc(base + 'npy')
serial.save(base + 'pkl', d)
