def _get_imagenet_weights_path(self):
if (not self.kernel_quantizer == "ste_sign" and
(self.input_quantizer in [
"ste_sign", "ste_unsign", "ste_shift_sign", "ste_shift_unsign"
])):
raise ValueError(
f"{self.model_name} only has ImageNet weights for the BNN variant"
)
if self.include_top:
weights_path = utils.download_pretrained_model(
model="r2b",
version="v0.1.0",
file="r2b_weights.h5",
file_hash=
"e8fd16ca1ab9810ac3835f24f5c62758a57bc32a615f73aaa50d382d2b9617e1",
)
else:
weights_path = utils.download_pretrained_model(
model="r2b",
version="v0.1.0",
file="r2b_weights_notop.h5",
file_hash=
"4ec47abf1a4da5c65f4908076257e8d5c812673891089a88c9d9e84e949d1dab",
)
return weights_path
def build(self) -> tf.keras.models.Model:
model = super().build()
# Load weights.
if self.weights == "imagenet":
# Download appropriate file
if self.include_top:
weights_path = utils.download_pretrained_model(
model="quicknet_large",
version="v0.2.0",
file="quicknet_large_weights.h5",
file_hash=
"2d9ebbf8ba0500552e4dd243c3e52fd8291f965ef6a0e1dbba13cc72bf6eee8b",
)
else:
weights_path = utils.download_pretrained_model(
model="quicknet_large",
version="v0.2.0",
file="quicknet_large_weights_notop.h5",
file_hash=
"067655ef8a1a1e99ef1c71fa775c09aca44bdfad0b9b71538b4ec500c3beee4f",
)
model.load_weights(weights_path)
elif self.weights is not None:
model.load_weights(self.weights)
return model
def build(self) -> tf.keras.models.Model:
model = super().build()
# Load weights.
if self.weights == "imagenet":
# Download appropriate file
if self.include_top:
weights_path = utils.download_pretrained_model(
model="quicknet_xl",
version="v0.1.0",
file="quicknet_xl_weights.h5",
file_hash=
"a85eea1204fa9a8401f922f94531858493e3518e3374347978ed7ba615410498",
)
else:
weights_path = utils.download_pretrained_model(
model="quicknet_xl",
version="v0.1.0",
file="quicknet_xl_weights_notop.h5",
file_hash=
"b97074d6618acde4201d1f8676d32272d27743ddfe27c6c97e4516511ebb5008",
)
model.load_weights(weights_path)
elif self.weights is not None:
model.load_weights(self.weights)
return model
def build(self) -> tf.keras.models.Model:
model = super().build()
# Load weights.
if self.weights == "imagenet":
# Download appropriate file
if self.include_top:
weights_path = utils.download_pretrained_model(
model="quicknet",
version="v0.2.0",
file="quicknet_weights.h5",
file_hash=
"6a765f120ba7b62a7740e842c4f462eb7ba3dd65eb46b4694c5bc8169618fae7",
)
else:
weights_path = utils.download_pretrained_model(
model="quicknet",
version="v0.2.0",
file="quicknet_weights_notop.h5",
file_hash=
"5bf2fc450fb8cc322b33a16410bf88fed09d05c221550c2d5805a04985383ac2",
)
model.load_weights(weights_path)
elif self.weights is not None:
model.load_weights(self.weights)
return model
def build(self) -> tf.keras.models.Model:
# Layer 1
out = tf.keras.layers.Conv2D(
self.filters,
(7, 7),
strides=2,
kernel_initializer=self.kernel_initializer,
padding="same",
use_bias=False,
)(self.image_input)
out = tf.keras.layers.BatchNormalization(momentum=0.8)(out)
out = tf.keras.layers.MaxPool2D((3, 3), strides=2, padding="same")(out)
# Layer 2
out = self.residual_block(out, filters=self.filters)
# Layer 3 - 5
for _ in range(3):
out = self.residual_block(out)
# Layer 6 - 17
for _ in range(3):
out = self.residual_block(out, double_filters=True)
for _ in range(3):
out = self.residual_block(out)
# Layer 18
if self.include_top:
out = utils.global_pool(out)
out = tf.keras.layers.Dense(self.num_classes)(out)
out = tf.keras.layers.Activation("softmax", dtype="float32")(out)
model = tf.keras.Model(inputs=self.image_input,
outputs=out,
name="birealnet18")
# Load weights.
if self.weights == "imagenet":
# Download appropriate file
if self.include_top:
weights_path = utils.download_pretrained_model(
model="birealnet",
version="v0.3.0",
file="birealnet_weights.h5",
file_hash=
"6e6efac1584fcd60dd024198c87f42eb53b5ec719a5ca1f527e1fe7e8b997117",
)
else:
weights_path = utils.download_pretrained_model(
model="birealnet",
version="v0.3.0",
file="birealnet_weights_notop.h5",
file_hash=
"5148b61c0c2a1094bdef811f68bf4957d5ba5f83ad26437b7a4a6855441ab46b",
)
model.load_weights(weights_path)
elif self.weights is not None:
model.load_weights(self.weights)
return model
def build(self) -> tf.keras.models.Model:
# Feature extractor
out = self.conv_block(
self.image_input,
features=64,
kernel_size=11,
strides=4,
pool=True,
first_layer=True,
)
out = self.conv_block(out, features=192, kernel_size=5, pool=True)
out = self.conv_block(out, features=384, kernel_size=3)
out = self.conv_block(out, features=384, kernel_size=3)
out = self.conv_block(
out, features=256, kernel_size=3, pool=True, no_inflation=True
)
# Classifier
if self.include_top:
try:
channels = out.shape[-1] * out.shape[-2] * out.shape[-3]
except TypeError:
channels = -1
out = tf.keras.layers.Reshape((1, 1, channels))(out)
out = self.dense_block(out, units=4096)
out = self.dense_block(out, units=4096)
out = self.dense_block(out, self.num_classes)
out = tf.keras.layers.Flatten()(out)
out = tf.keras.layers.Activation("softmax", dtype="float32")(out)
model = tf.keras.models.Model(
inputs=self.image_input, outputs=out, name="binary_alexnet"
)
# Load weights.
if self.weights == "imagenet":
# Download appropriate file
if self.include_top:
weights_path = utils.download_pretrained_model(
model="binary_alexnet",
version="v0.3.0",
file="binary_alexnet_weights.h5",
file_hash="7fc065c47c5c1d92389e0bb988ce6df6a4fa09d803b866e2ba648069d6652d63",
)
else:
weights_path = utils.download_pretrained_model(
model="binary_alexnet",
version="v0.2.1",
file="binary_alexnet_weights_notop.h5",
file_hash="1d41b33ff39cd28d13679392641bf7711174a96d182417f91df45d0548f5bb47",
)
model.load_weights(weights_path)
elif self.weights is not None:
model.load_weights(self.weights)
return model
def build(self) -> tf.keras.models.Model:
# Feature extractor
out = self.conv_block(
self.image_input,
features=64,
kernel_size=11,
strides=4,
pool=True,
first_layer=True,
)
out = self.conv_block(out, features=192, kernel_size=5, pool=True)
out = self.conv_block(out, features=384, kernel_size=3)
out = self.conv_block(out, features=384, kernel_size=3)
out = self.conv_block(
out, features=256, kernel_size=3, pool=True, no_inflation=True
)
# Classifier
if self.include_top:
out = tf.keras.layers.Flatten()(out)
out = self.dense_block(out, units=4096)
out = self.dense_block(out, units=4096)
out = self.dense_block(out, self.num_classes)
out = tf.keras.layers.Activation("softmax", dtype="float32")(out)
model = tf.keras.models.Model(
inputs=self.image_input, outputs=out, name="binary_alexnet"
)
# Load weights.
if self.weights == "imagenet":
# Download appropriate file
if self.include_top:
weights_path = utils.download_pretrained_model(
model="binary_alexnet",
version="v0.2.0",
file="binary_alexnet_weights.h5",
file_hash="0f8d3f6c1073ef993e2e99a38f8e661e5efe385085b2a84b43a7f2af8500a3d3",
)
else:
weights_path = utils.download_pretrained_model(
model="binary_alexnet",
version="v0.2.1",
file="binary_alexnet_weights_notop.h5",
file_hash="1d41b33ff39cd28d13679392641bf7711174a96d182417f91df45d0548f5bb47",
)
model.load_weights(weights_path)
elif self.weights is not None:
model.load_weights(self.weights)
return model
def build(self) -> tf.keras.models.Model:
out = tf.keras.layers.Conv2D(96,
kernel_size=12,
strides=4,
padding="valid",
use_bias=True)(self.image_input)
out = self.conv_block(out, filters=256, kernel_size=5, pool=True)
out = self.conv_block(out, filters=384, kernel_size=3, pool=True)
out = self.conv_block(out, filters=384, kernel_size=3)
out = self.conv_block(out,
filters=256,
kernel_size=3,
pool_padding="valid",
pool=True)
if self.include_top:
out = tf.keras.layers.Flatten()(out)
out = self.fully_connected_block(out, units=4096)
out = self.fully_connected_block(out, units=4096)
out = tf.keras.layers.Activation("clip_by_value_activation")(out)
out = tf.keras.layers.Dense(self.num_classes, use_bias=True)(out)
out = tf.keras.layers.Activation("softmax", dtype="float32")(out)
model = tf.keras.Model(inputs=self.image_input,
outputs=out,
name="dorefanet")
# Load weights.
if self.weights == "imagenet":
# Download appropriate file
if self.include_top:
weights_path = utils.download_pretrained_model(
model="dorefanet",
version="v0.1.0",
file="dorefanet_weights.h5",
file_hash=
"645d7839d574faa3eeeca28f3115773d75da3ab67ff6876b4de12d10245ecf6a",
)
else:
weights_path = utils.download_pretrained_model(
model="dorefanet",
version="v0.1.0",
file="dorefanet_weights_notop.h5",
file_hash=
"679368128e19a2a181bfe06ca3a3dec368b1fd8011d5f42647fbbf5a7f36d45f",
)
model.load_weights(weights_path)
elif self.weights is not None:
model.load_weights(self.weights)
return model
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_37_dilated_weights.h5",
file_hash="15c1bcd79b8dc22971382fbf79acf364a3f51049d0e584a11533e6fdbb7363d3",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_45_weights_notop.h5",
file_hash="e72d5cc6b0afe4612f8be7b1f9bb48a53ba2c8468b57bf1266d2900c99fd2adf",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_37_weights.h5",
file_hash="8056a5d52c3ed86a934893987d09a06f59a5166aa9bddcaedb050f111d0a7d76",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="meliusnet22",
version="v0.1.0",
file="meliusnet22_weights.h5",
file_hash="c1ba85e8389ae326009665ec13331e49fc3df4d0f925fa8553e224f7362c18ed",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.1",
file="binary_densenet_37_dilated_weights_notop.h5",
file_hash="8b31fbfdc8de08a46c6adcda1ced48ace0a2ff0ce45a05c72b2acc27901dd88b",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_28_weights_notop.h5",
file_hash="a376df1e41772c4427edd1856072b934a89bf293bf911438bf6f751a9b2a28f5",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_28_weights.h5",
file_hash="21fe3ca03eed244df9c41a2219876fcf03e73800932ec96a3e2a76af4747ac53",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_45_weights.h5",
file_hash="d00a0d26fbd2dba1bfba8c0306c770f3aeea5c370e99f963bb239bd916f72c37",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_37_weights_notop.h5",
file_hash="4e12bca9fd27580a5b833241c4eb35d6cc332878c406048e6ca8dbbc78d59175",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="meliusnet22",
version="v0.1.1",
file="meliusnet22_weights_notop.h5",
file_hash="abfc5c50049d72a14e44df0c1cb73896ece2a1ab4bf9bb48fede6cc2f5e0b58f",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="quicknet_xl",
version="v0.1.1",
file="quicknet_xl_weights_notop.h5",
file_hash=
"ad5cbfa333b0aabde75dc524c9ce4a5ae096061da0e2dcf362ec6e587a83a511",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="quicknet_xl",
version="v0.1.1",
file="quicknet_xl_weights.h5",
file_hash=
"19a41e753dbd4fbc3cbdaecd3627fb536ef55d64702996aae3875a8de3cf8073",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="quicknet_large",
version="v0.2.1",
file="quicknet_large_weights_notop.h5",
file_hash=
"b65d59dd2d5af63d019997b05faff9e003510e2512aa973ee05eb1b82b8792a9",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="quicknet_large",
version="v0.2.1",
file="quicknet_large_weights.h5",
file_hash=
"6bf778e243466c678d6da0e3a91c77deec4832460046fca9e6ac8ae97a41299c",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v0.2.1",
file="quicknet_weights_notop.h5",
file_hash=
"359eed6dae43525eddf520ea87ec9b54750ee0e022647775d115a38856be396f",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v1.0",
file="quicknet_weights_notop.h5",
file_hash=
"204414e438373f14f6056a1c098249f505a87dd238e18d3a47a9bd8b66227881",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v1.0",
file="quicknet_weights.h5",
file_hash=
"8aba9e4e5f8d342faef04a0b2ae8e562da57dbb7d15162e8b3e091c951ba756c",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v1.0",
file="quicknet_large_weights.h5",
file_hash=
"c5158e8a59147b31370becd937825f4db8a5cdf308314874f678f596629be45c",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v1.0",
file="quicknet_small_weights_notop.h5",
file_hash=
"be8ba657155846be355c5580d1ea56eaf8282616de065ffc39257202f9f164ea",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v1.0",
file="quicknet_small_weights.h5",
file_hash=
"1ac3b07df7f5a911dd0b49febb2486428ddf1ca130297c403815dfae5a1c71a2",
)
def imagenet_no_top_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v1.0",
file="quicknet_large_weights_notop.h5",
file_hash=
"adcf154a2a8007e81bd6af77c035ffbf54cd6413b89a0ba294e23e76a82a1b78",
)
def imagenet_weights_path(self):
return utils.download_pretrained_model(
model="quicknet",
version="v0.2.1",
file="quicknet_weights.h5",
file_hash=
"7b4fa94f5241c7aad3412ca42b5db6517dbc4847cff710cb82be10c2f83bc0be",
)
