def DoReFaNet(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the DoReFa-net architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
dorefanet-v0.1.0/dorefanet.json
```
```plot-altair
/plots/dorefanet.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low
Bitwidth Gradients](https://arxiv.org/abs/1606.06160)
"""
input_shape = utils.validate_input(input_shape, weights, include_top,
classes)
model = dorefa_net(
default(),
input_shape,
classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if 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 weights is not None:
model.load_weights(weights)
return model
def BinaryAlexNet(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the BinaryAlexNet architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
binary_alexnet-v0.2.0/binary_alexnet.json
```
```plot-altair
/plots/binary_alexnet.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
"""
input_shape = utils.validate_input(input_shape, weights, include_top,
classes)
model = binary_alexnet(
default(),
input_shape,
classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if 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="v0.2.0",
file="binary_alexnet_weights_notop.h5",
file_hash=
"1c7e2ef156edd8e7615e75a3b8929f9025279a948d1911824c2f5a798042475e",
)
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def BinaryDenseNet45(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the Binary BinaryDenseNet 45 architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
binary_densenet-v0.1.0/binary_densenet_45.json
```
```plot-altair
/plots/densenet_45.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [Back to Simplicity:
How to Train Accurate BNNs from Scratch?](https://arxiv.org/abs/1906.08637)
"""
input_shape = utils.validate_input(input_shape, weights, include_top,
classes)
model = binary_densenet(
binary_densenet45(),
input_shape=input_shape,
num_classes=classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if include_top:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_45_weights.h5",
file_hash=
"d00a0d26fbd2dba1bfba8c0306c770f3aeea5c370e99f963bb239bd916f72c37",
)
else:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_45_weights_notop.h5",
file_hash=
"e72d5cc6b0afe4612f8be7b1f9bb48a53ba2c8468b57bf1266d2900c99fd2adf",
)
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def BiRealNet(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the Bi-Real Net architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
birealnet-v0.3.0/birealnet.json
```
```plot-altair
/plots/birealnet.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [Bi-Real Net: Enhancing the Performance of 1-bit CNNs With Improved
Representational Capability and Advanced Training
Algorithm](https://arxiv.org/abs/1808.00278)
"""
input_shape = utils.validate_input(input_shape, weights, include_top, classes)
model = birealnet(
default(),
input_shape,
classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if 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 weights is not None:
model.load_weights(weights)
return model
def BinaryDenseNet37Dilated(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the Dilated BinaryDenseNet 37 architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
binary_densenet-v0.1.0/binary_densenet_37_dilated.json
```
```plot-altair
/plots/densenet_37_dilated.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [Back to Simplicity:
How to Train Accurate BNNs from Scratch?](https://arxiv.org/abs/1906.08637)
"""
input_shape = utils.validate_input(input_shape, weights, include_top,
classes)
model = binary_densenet(
binary_densenet37_dilated(),
input_shape=input_shape,
num_classes=classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if include_top:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_37_dilated_weights.h5",
file_hash=
"15c1bcd79b8dc22971382fbf79acf364a3f51049d0e584a11533e6fdbb7363d3",
)
else:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_37_dilated_weights_notop.h5",
file_hash=
"eaf3eac19fc90708f56a27435fb06d0e8aef40e6e0411ff7a8eefbe479226e4f",
)
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def BinaryDenseNet37(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the BinaryDenseNet 37 architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
binary_densenet-v0.1.0/binary_densenet_37.json
```
```plot-altair
/plots/densenet_37.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [Back to Simplicity:
How to Train Accurate BNNs from Scratch?](https://arxiv.org/abs/1906.08637)
"""
input_shape = utils.validate_input(input_shape, weights, include_top,
classes)
model = binary_densenet(
binary_densenet37(),
input_shape=input_shape,
num_classes=classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if include_top:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_37_weights.h5",
file_hash=
"8056a5d52c3ed86a934893987d09a06f59a5166aa9bddcaedb050f111d0a7d76",
)
else:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_37_weights_notop.h5",
file_hash=
"4e12bca9fd27580a5b833241c4eb35d6cc332878c406048e6ca8dbbc78d59175",
)
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def BinaryDenseNet28(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the BinaryDenseNet 28 architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
binary_densenet-v0.1.0/binary_densenet_28.json
```
```plot-altair
/plots/densenet_28.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [Back to Simplicity:
How to Train Accurate BNNs from Scratch?](https://arxiv.org/abs/1906.08637)
"""
input_shape = utils.validate_input(input_shape, weights, include_top,
classes)
model = binary_densenet(
binary_densenet28(),
input_shape=input_shape,
num_classes=classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if include_top:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_28_weights.h5",
file_hash=
"21fe3ca03eed244df9c41a2219876fcf03e73800932ec96a3e2a76af4747ac53",
)
else:
weights_path = utils.download_pretrained_model(
model="binary_densenet",
version="v0.1.0",
file="binary_densenet_28_weights_notop.h5",
file_hash=
"a376df1e41772c4427edd1856072b934a89bf293bf911438bf6f751a9b2a28f5",
)
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def XNORNet(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the XNOR-Net architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
xnornet-v0.2.0/xnornet.json
```
```plot-altair
/plots/xnornet.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False
(otherwise the input shape has to be `(224, 224, 3)` (with `channels_last` data
format) or `(3, 224, 224)` (with `channels_first` data format).
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [XNOR-Net: ImageNet Classification Using Binary Convolutional Neural
Networks](https://arxiv.org/abs/1603.05279)
"""
input_shape = utils.validate_input(input_shape, weights, include_top,
classes)
model = xnornet(
default(),
input_shape,
classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if include_top:
weights_path = utils.download_pretrained_model(
model="xnornet",
version="v0.2.0",
file="xnornet_weights.h5",
file_hash=
"e6ba24f785655260ae76a2ef1fab520e3528243d9c8fac430299cd81dbeabe10",
)
else:
weights_path = utils.download_pretrained_model(
model="xnornet",
version="v0.2.0",
file="xnornet_weights_notop.h5",
file_hash=
"0b8e3d0d60a7a728b5e387b8cd9f0fedc1dd72bcf9f4c693a2245d3a3c596b91",
)
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def BinaryResNetE18(
include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
classes=1000,
):
"""Instantiates the BinaryResNetE 18 architecture.
Optionally loads weights pre-trained on ImageNet.
```netron
resnet_e-v0.1.0/resnet_e_18.json
```
```plot-altair
/plots/resnet_e_18.vg.json
```
# Arguments
include_top: whether to include the fully-connected layer at the top of the network.
weights: one of `None` (random initialization), "imagenet" (pre-training on
ImageNet), or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as
image input for the model.
input_shape: optional shape tuple, only to be specified if `include_top` is False,
otherwise the input shape has to be `(224, 224, 3)`.
It should have exactly 3 inputs channels.
classes: optional number of classes to classify images into, only to be specified
if `include_top` is True, and if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`, or invalid input shape.
# References
- [Back to Simplicity:
How to Train Accurate BNNs from Scratch?](https://arxiv.org/abs/1906.08637)
"""
input_shape = utils.validate_input(input_shape, weights, include_top, classes)
model = resnet_e(
default(),
input_shape=input_shape,
num_classes=classes,
input_tensor=input_tensor,
include_top=include_top,
)
# Load weights.
if weights == "imagenet":
# download appropriate file
if include_top:
weights_path = utils.download_pretrained_model(
model="resnet_e",
version="v0.1.0",
file="resnet_e_18_weights.h5",
file_hash="bde4a64d42c164a7b10a28debbe1ad5b287c499bc0247ecb00449e6e89f3bf5b",
)
else:
weights_path = utils.download_pretrained_model(
model="resnet_e",
version="v0.1.0",
file="resnet_e_18_weights_notop.h5",
file_hash="14cb037e47d223827a8d09db88ec73d60e4153a4464dca847e5ae1a155e7f525",
)
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model