embeddings_initializer='glorot_normal')
emb = Flatten()(lookup_table(image_class))
# hadamard product between z-space and a class conditional embedding
hc = multiply([latent, emb])
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([hc, scale(input_energy, 100)])
generator_inputs.append(image_class)
else:
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([latent, scale(input_energy, 100)])
# each of these builds a LAGAN-inspired [arXiv/1701.05927] component with
# linear last layer
img_layer0 = build_generator(h, 3, 96)
img_layer1 = build_generator(h, 12, 12)
img_layer2 = build_generator(h, 12, 6)
if not no_attn:
logger.info('using attentional mechanism')
# resizes from (3, 96) => (12, 12)
zero2one = AveragePooling2D(pool_size=(1, 8))(
UpSampling2D(size=(4, 1))(img_layer0))
img_layer1 = inpainting_attention(img_layer1, zero2one)
# resizes from (12, 12) => (12, 6)
one2two = AveragePooling2D(pool_size=(1, 2))(img_layer1)
img_layer2 = inpainting_attention(img_layer2, one2two)
emb = Flatten()(lookup_table(image_class))
# hadamard product between z-space and a class conditional embedding
hc = multiply([latent, emb])
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([hc, scale(input_energy, 100)])
generator_inputs.append(image_class)
else:
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([latent, scale(input_energy, 100)])
# each of these builds a LAGAN-inspired [arXiv/1701.05927] component
img_layer = []
for i in range(5):
img_layer.append(build_generator(h, 12, 12))
if not no_attn:
logger.info('using attentional mechanism')
zero2one = AveragePooling2D(pool_size=(1, 1))(
UpSampling2D(size=(1, 1))(img_layer[0]))
img_layer[1] = inpainting_attention(img_layer[1], zero2one)
for j in range(1, 4):
one2N = AveragePooling2D(pool_size=(1, 1))(img_layer[j])
img_layer[j + 1] = inpainting_attention(img_layer[j + 1], one2N)
generator_outputs = []
for k in range(5):
generator_outputs.append(Activation('relu')(img_layer[k]))
emb = Flatten()(lookup_table(image_class))
# hadamard product between z-space and a class conditional embedding
hc = multiply([latent, emb])
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([hc, scale(input_energy, 100)])
generator_inputs.append(image_class)
else:
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([latent, scale(input_energy, 100)])
# each of these builds a LAGAN-inspired [arXiv/1701.05927] component
img_layer = []
for i in range(5):
img_layer.append(build_generator(h, 20, 20))
if not no_attn:
logger.info('using attentional mechanism')
zero2one = AveragePooling2D(pool_size=(1, 1))(
UpSampling2D(size=(1, 1))(img_layer[0]))
img_layer[1] = inpainting_attention(img_layer[1], zero2one)
for j in range(1, 4):
one2N = AveragePooling2D(pool_size=(1, 1))(img_layer[j])
img_layer[j + 1] = inpainting_attention(img_layer[j + 1], one2N)
generator_outputs = []
for k in range(5):
generator_outputs.append(Activation('relu')(img_layer[k]))
embeddings_initializer='glorot_normal')
emb = Flatten()(lookup_table(image_class))
# hadamard product between z-space and a class conditional embedding
hc = multiply([latent, emb])
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([hc, scale(input_energy, 100)])
generator_inputs.append(image_class)
else:
# requested energy comes in GeV
h = Lambda(lambda x: x[0] * x[1])([latent, scale(input_energy, 100)])
# each of these builds a LAGAN-inspired [arXiv/1701.05927] component with
# linear last layer
img_layer0 = build_generator(h, 20, 20)
img_layer1 = build_generator(h, 20, 20)
img_layer2 = build_generator(h, 20, 20)
if not no_attn:
logger.info('using attentional mechanism')
# resizes from (20, 20) => (20, 20)
zero2one = AveragePooling2D(pool_size=(1, 1))(
UpSampling2D(size=(1, 1))(img_layer0))
img_layer1 = inpainting_attention(img_layer1, zero2one)
# resizes from (20, 20) => (20, 20)
one2two = AveragePooling2D(pool_size=(1, 1))(img_layer1)
img_layer2 = inpainting_attention(img_layer2, one2two)
embeddings_initializer='glorot_normal')(image_class))
# hadamard product between z-space and a class conditional embedding
hc = merge([latent, emb], mode='mul')
h = Lambda(lambda x: x[0] * x[1])([hc, scale(input_energy, 100)])
generator_inputs.append(image_class)
else:
h = Lambda(lambda x: x[0] * x[1])([latent, scale(input_energy, 100)])
# h = concatenate([latent, input_energy])
# emb = Flatten()(Embedding(nb_classes, latent_size, input_length=1,
# embeddings_initializer='glorot_normal')(image_class))
# # hadamard product between z-space and a class conditional embedding
# h = merge([latent, emb], mode='mul')
img_layer0 = build_generator(h, 5, 5)
img_layer1 = build_generator(h, 5, 5)
# img_layer2 = build_generator(h, 12, 6)
# if parse_args.in_paint:
#
# # inpainting
# zero2one = AveragePooling2D(pool_size=(1, 8))(
# UpSampling2D(size=(4, 1))(img_layer0))
# # final_img_layer1 = add([zero2one, img_layer1])
# img_layer1 = inpainting_attention(img_layer1, zero2one)
#
# one2two = AveragePooling2D(pool_size=(1, 2))(img_layer1)
# # final_img_layer2 = add([one2two, img_layer2])
# img_layer2 = inpainting_attention(img_layer2, one2two)