def run():
if True:
download_pretrained_model.run()
if True:
print "Bulid reconstructor"
build_reconstructor.run(
model_name="alexnet",
layer_names=["data", "conv1", "conv2", "conv3", "conv4", "conv5"],
layer_target=["conv1", "conv2", "conv3", "conv4", "conv5"],
image_size=(227, 227),
target_layer="fc8",
)
if True:
print "Test models"
reconstruct.run("./testdata/024_227.png", "reconstructed.png")
compute_saliency_map.run("./testdata/024.jpg", "saliency_map.png")
def run():
if True:
download_pretrained_model.run()
if True:
print 'Bulid reconstructor'
build_reconstructor.run(
model_name='alexnet',
layer_names=['data', 'conv1', 'conv2', 'conv3', 'conv4', 'conv5'],
layer_target=[
'conv1',
'conv2',
'conv3',
'conv4',
'conv5',
],
image_size=(227, 227),
target_layer='fc8',
)
if True:
print 'Test models'
reconstruct.run('./testdata/024_227.png', 'reconstructed.png')
compute_saliency_map.run('./testdata/024.jpg', 'saliency_map.png')
def run():
if False:
print 'build theano models'
build_alexnet.run()
build_vggnet.run()
if False:
print 'prepare images'
prepare_images.run()
if False:
print 'compute mean and std of feature maps'
compute_mean_std.run(
model_name='alexnet',
layer_names=[
'data',
'conv1',
'conv2',
'conv3',
'conv4',
'conv5',
'fc6', 'fc7', 'fc8',
],
image_size=(227, 227),
)
compute_mean_std.run(
model_name='vggnet',
layer_names=[
'data',
'conv1_1', 'conv1_2',
'conv2_1', 'conv2_2',
'conv3_1', 'conv3_2', 'conv3_3', 'conv3_4',
'conv4_1', 'conv4_2', 'conv4_3', 'conv4_4',
'conv5_1', 'conv5_2', 'conv5_3', 'conv5_4',
'fc6', 'fc7', 'fc8'
],
image_size=(224, 224),
)
if False:
print 'PCA of feature maps'
train_pca.run(
model_name='alexnet',
layer_names=[
'data',
'conv1',
'conv2',
'conv3',
'conv4',
'conv5'
],
image_size=(227, 227),
)
train_pca.run(
model_name='vggnet',
layer_names=[
'data',
'conv1_1', 'conv1_2',
'conv2_1', 'conv2_2',
'conv3_1', 'conv3_2', 'conv3_3', 'conv3_4',
'conv4_1', 'conv4_2', 'conv4_3', 'conv4_4',
'conv5_1', 'conv5_2', 'conv5_3', 'conv5_4',
],
image_size=(224, 224),
)
if False:
print 'Train language model'
train_vlm.run(
model_name='alexnet',
layer_names=['data', 'conv1', 'conv2', 'conv3', 'conv4', 'conv5'],
layer_sizes=[3, 96, 256, 384, 384, 256],
image_size=(227, 227),
)
train_vlm.run(
model_name='vggnet',
layer_names=[
'data',
'conv1_1',
'conv2_1',
'conv3_1',
'conv4_1',
'conv5_1',
],
layer_sizes=[3, 64, 128, 256, 512, 512],
batch_size=1,
lr=2e+1,
image_size=(224, 224),
)
train_vlm.run(
model_name='vggnet',
layer_names=[
'data',
'conv1_2',
'conv2_2',
'conv3_4',
'conv4_4',
'conv5_4',
],
layer_sizes=[3, 64, 128, 256, 512, 512],
batch_size=1,
lr=2e+1,
image_size=(224, 224),
)
if False:
print 'Bulid reconstructor'
build_reconstructor.run(
model_name='alexnet',
layer_names=[
'data',
'conv1',
'conv2',
'conv3',
'conv4',
'conv5'
],
layer_target=[
'conv1',
'conv2',
'conv3',
'conv4',
'conv5',
],
image_size=(227, 227),
target_layer='fc8',
)
if True:
print 'Test models'
reconstruct.run('./testdata/024_227.png', 'reconstructed.png')
compute_saliency_map.run('./testdata/024.jpg', 'saliency_map.png')