pnet.PartsLayer(
numParts,
(patchSize, patchSize),
settings=dict(
outer_frame=0,
#em_seed=training_seed,
threshold=40,
samples_per_image=40,
max_samples=1000000,
min_prob=0.005,
)),
pnet.PoolingLayer(shape=(4, 4), strides=(4, 4)),
pnet.SVMClassificationLayer(C=None)
]
net = pnet.PartsNet(layers)
digits = range(10)
print('Extracting subsets...')
ims10k = data[:10000]
print('Done.')
start0 = time.time()
print('Training unsupervised...')
net.train(ims10k)
print('Done.')
end0 = time.time()
error_rates = []
test_ims, test_labels = ag.io.load_mnist('testing', return_labels=True)
for n_classes in num_class_models0:
print('Classifier:', classifier, 'Components:', n_classes)
if classifier == 'mixture':
layers = [
pnet.PoolingLayer(shape=(4, 4), strides=(4, 4)),
pnet.MixtureClassificationLayer(n_components=n_classes,
min_prob=1e-5)
]
elif classifier == 'svm':
layers = [
pnet.PoolingLayer(shape=(4, 4), strides=(4, 4)),
pnet.SVMClassificationLayer(C=None),
]
clnet = pnet.PartsNet([net] + layers)
start1 = time.time()
print('Training supervised...')
print(sup_ims.shape)
clnet.train(sup_ims, sup_labels)
print('Done.')
end1 = time.time()
corrects = 0
total = 0
if 0:
test_ims, test_labels = mnist_data['test_image'], mnist_data[
'test_label']
else:
test_ims = ims2k
print("Inside")
layers = [
pnet.EdgeLayer(k=5, radius=1, spread='orthogonal', minimum_contrast=0.05),#
pnet.PartsLayer(numParts, (patchSize, patchSize), settings=dict(outer_frame=0,
#em_seed=training_seed,
threshold=40,
samples_per_image=40,
max_samples=1000000,
min_prob=0.005,
)),
]
net = pnet.PartsNet(layers)
digits = range(10)
print('Extracting subsets...')
ims10k = data[:10000]
print('Done.')
start0 = time.time()
print('Training unsupervised...')
net.train(ims10k)
print('Done.')
end0 = time.time()
error_rates = []
extensionPatchSize = args.extensionPatchSize
data = np.load(dataFileName)
training_seed = args.seed
num_rotation = args.num_rotation
net = pnet.PartsNet.load(args.model)
print("Inside")
extensionlayers = [
pnet.RotExtensionPartsLayer(num_parts=numParts,
num_components=numExtensionParts,
rotation=num_rotation,
part_shape=(extensionPatchSize,
extensionPatchSize),
lowerLayerShape=(6, 6))
]
clnet = pnet.PartsNet([net] + extensionlayers)
digits = range(10)
print('Extracting subsets...')
ims10k = data[:10000]
print('Done.')
start0 = time.time()
print('Training unsupervised...')
clnet.train(ims10k)
print('Done.')
end0 = time.time()
clnet.save(saveFile)
print("=================")
print(sup_ims.shape)
print(sup_labels)
for classifier in 'mixture', 'svm':
for rotspread in [0, 1]:
net.layers[0]._settings['rotation_spreading_radius'] = rotspread
print('Classifier:', classifier, 'Rotational spreading:', rotspread)
if classifier == 'mixture':
cl = pnet.MixtureClassificationLayer(n_components=numOfClassModel, min_prob=1e-5)
elif classifier == 'svm':
cl = pnet.SVMClassificationLayer(C=None)
clnet = pnet.PartsNet([net, cl])
start1 = time.time()
print('Training supervised...')
print(sup_ims.shape)
clnet.train(sup_ims, sup_labels)
print('Done.')
end1 = time.time()
#print("Now testing...")
### Test ######################################################################
corrects = 0
total = 0
if 0:
test_ims, test_labels = mnist_data['test_image'], mnist_data['test_label']
