Further style fixes in mobilenet
Esse commit está contido em:
Francois Chollet
@@ -3,40 +3,47 @@
|
||||
MobileNet is a general architecture and can be used for multiple use cases.
|
||||
Depending on the use case, it can use different input layer size and
|
||||
different width factors. This allows different width models to reduce
|
||||
the number of multiply-adds and thereby reduce inference cost on mobile devices.
|
||||
the number of multiply-adds and thereby
|
||||
reduce inference cost on mobile devices.
|
||||
|
||||
MobileNets support any input size greater than 32 x 32, with larger image sizes
|
||||
offering better performance. The number of parameters and number of multiply-adds
|
||||
can be modified by using the `alpha` parameter, which increases/decreases the number
|
||||
of filters in each layer. By altering the image size and `alpha` parameter,
|
||||
offering better performance.
|
||||
The number of parameters and number of multiply-adds
|
||||
can be modified by using the `alpha` parameter,
|
||||
which increases/decreases the number of filters in each layer.
|
||||
By altering the image size and `alpha` parameter,
|
||||
all 16 models from the paper can be built, with ImageNet weights provided.
|
||||
|
||||
The paper demonstrates the performance of MobileNets using `alpha` values of
|
||||
1.0 (also called 100 % MobileNet), 0.75, 0.5 and 0.25. For each of these `alpha`
|
||||
values, weights for 4 different input image sizes are provided (224, 192, 160, 128).
|
||||
1.0 (also called 100 % MobileNet), 0.75, 0.5 and 0.25.
|
||||
For each of these `alpha` values, weights for 4 different input image sizes
|
||||
are provided (224, 192, 160, 128).
|
||||
|
||||
The following table describes the size and accuracy of the 100% MobileNet on size 224 x 224
|
||||
-----------------------------------------------------------------------------------------
|
||||
Width Multiplier (alpha) | ImageNet Accuracy | Million Multiply-Adds | Million Parameters
|
||||
-----------------------------------------------------------------------------------------
|
||||
| 1.0 MobileNet-224 | 70.6 % | 529 | 4.2 |
|
||||
| 0.75 MobileNet-224 | 68.4 % | 325 | 2.6 |
|
||||
| 0.50 MobileNet-224 | 63.7 % | 149 | 1.3 |
|
||||
| 0.25 MobileNet-224 | 50.6 % | 41 | 0.5 |
|
||||
-----------------------------------------------------------------------------------------
|
||||
The following table describes the size and accuracy of the 100% MobileNet
|
||||
on size 224 x 224:
|
||||
----------------------------------------------------------------------------
|
||||
Width Multiplier (alpha) | ImageNet Acc | Multiply-Adds (M) | Params (M)
|
||||
----------------------------------------------------------------------------
|
||||
| 1.0 MobileNet-224 | 70.6 % | 529 | 4.2 |
|
||||
| 0.75 MobileNet-224 | 68.4 % | 325 | 2.6 |
|
||||
| 0.50 MobileNet-224 | 63.7 % | 149 | 1.3 |
|
||||
| 0.25 MobileNet-224 | 50.6 % | 41 | 0.5 |
|
||||
----------------------------------------------------------------------------
|
||||
|
||||
The following table describes the performance of the 100 % MobileNet on various input sizes
|
||||
-----------------------------------------------------------------------------------------
|
||||
Resolution | ImageNet Accuracy | Million Multiply-Adds | Million Parameters
|
||||
-----------------------------------------------------------------------------------------
|
||||
| 1.0 MobileNet-224 | 70.6 % | 529 | 4.2 |
|
||||
| 1.0 MobileNet-192 | 69.1 % | 529 | 4.2 |
|
||||
| 1.0 MobileNet-160 | 67.2 % | 529 | 4.2 |
|
||||
| 1.0 MobileNet-128 | 64.4 % | 529 | 4.2 |
|
||||
-----------------------------------------------------------------------------------------
|
||||
The following table describes the performance of
|
||||
the 100 % MobileNet on various input sizes:
|
||||
------------------------------------------------------------------------
|
||||
Resolution | ImageNet Acc | Multiply-Adds (M) | Params (M)
|
||||
------------------------------------------------------------------------
|
||||
| 1.0 MobileNet-224 | 70.6 % | 529 | 4.2 |
|
||||
| 1.0 MobileNet-192 | 69.1 % | 529 | 4.2 |
|
||||
| 1.0 MobileNet-160 | 67.2 % | 529 | 4.2 |
|
||||
| 1.0 MobileNet-128 | 64.4 % | 529 | 4.2 |
|
||||
------------------------------------------------------------------------
|
||||
|
||||
The weights for all 16 models are obtained and translated from Tensorflow checkpoints
|
||||
found in https://github.com/tensorflow/models/blob/master/slim/nets/mobilenet_v1.md
|
||||
The weights for all 16 models are obtained and translated
|
||||
from Tensorflow checkpoints found at
|
||||
https://github.com/tensorflow/models/blob/master/slim/nets/mobilenet_v1.md
|
||||
|
||||
# Reference
|
||||
- [MobileNets: Efficient Convolutional Neural Networks for
|
||||
@@ -353,7 +360,8 @@ def MobileNet(input_shape=None,
|
||||
|
||||
if K.backend() != 'tensorflow':
|
||||
raise RuntimeError('Only Tensorflow backend is currently supported, '
|
||||
'as other backends do not support depthwise convolution.')
|
||||
'as other backends do not support '
|
||||
'depthwise convolution.')
|
||||
|
||||
if weights not in {'imagenet', None}:
|
||||
raise ValueError('The `weights` argument should be either '
|
||||
@@ -370,7 +378,8 @@ def MobileNet(input_shape=None,
|
||||
'depth multiplier must be 1')
|
||||
|
||||
if alpha not in [0.25, 0.50, 0.75, 1.0]:
|
||||
raise ValueError('If imagenet weights are being loaded, alpha can be one of'
|
||||
raise ValueError('If imagenet weights are being loaded, '
|
||||
'alpha can be one of'
|
||||
'`0.25`, `0.50`, `0.75` or `1.0` only.')
|
||||
|
||||
rows, cols = (0, 1) if K.image_data_format() == 'channels_last' else (1, 2)
|
||||
@@ -399,8 +408,8 @@ def MobileNet(input_shape=None,
|
||||
old_data_format = None
|
||||
|
||||
# Determine proper input shape. Note, include_top is False by default, as
|
||||
# input shape can be anything larger than 32x32 and the same number of parameters
|
||||
# will be used.
|
||||
# input shape can be anything larger than 32x32
|
||||
# and the same number of parameters will be used.
|
||||
input_shape = _obtain_input_shape(input_shape,
|
||||
default_size=224,
|
||||
min_size=32,
|
||||
@@ -416,23 +425,27 @@ def MobileNet(input_shape=None,
|
||||
img_input = input_tensor
|
||||
|
||||
x = _conv_block(img_input, 32, alpha, strides=(2, 2))
|
||||
x = _depthwise_conv_block(x, 64, alpha, depth_multiplier, id=1)
|
||||
x = _depthwise_conv_block(x, 64, alpha, depth_multiplier, block_id=1)
|
||||
|
||||
x = _depthwise_conv_block(x, 128, alpha, depth_multiplier, strides=(2, 2), id=2)
|
||||
x = _depthwise_conv_block(x, 128, alpha, depth_multiplier, id=3)
|
||||
x = _depthwise_conv_block(x, 128, alpha, depth_multiplier,
|
||||
strides=(2, 2), block_id=2)
|
||||
x = _depthwise_conv_block(x, 128, alpha, depth_multiplier, block_id=3)
|
||||
|
||||
x = _depthwise_conv_block(x, 256, alpha, depth_multiplier, strides=(2, 2), id=4)
|
||||
x = _depthwise_conv_block(x, 256, alpha, depth_multiplier, id=5)
|
||||
x = _depthwise_conv_block(x, 256, alpha, depth_multiplier,
|
||||
strides=(2, 2), block_id=4)
|
||||
x = _depthwise_conv_block(x, 256, alpha, depth_multiplier, block_id=5)
|
||||
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, strides=(2, 2), id=6)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, id=7)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, id=8)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, id=9)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, id=10)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, id=11)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier,
|
||||
strides=(2, 2), block_id=6)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=7)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=8)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=9)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=10)
|
||||
x = _depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=11)
|
||||
|
||||
x = _depthwise_conv_block(x, 1024, alpha, depth_multiplier, strides=(2, 2), id=12)
|
||||
x = _depthwise_conv_block(x, 1024, alpha, depth_multiplier, id=13)
|
||||
x = _depthwise_conv_block(x, 1024, alpha, depth_multiplier,
|
||||
strides=(2, 2), block_id=12)
|
||||
x = _depthwise_conv_block(x, 1024, alpha, depth_multiplier, block_id=13)
|
||||
|
||||
if include_top:
|
||||
if K.image_data_format() == 'channels_first':
|
||||
@@ -443,7 +456,8 @@ def MobileNet(input_shape=None,
|
||||
x = GlobalAveragePooling2D()(x)
|
||||
x = Reshape(shape, name='reshape_1')(x)
|
||||
x = Dropout(dropout, name='dropout')(x)
|
||||
x = Convolution2D(classes, (1, 1), padding='same', name='conv_preds')(x)
|
||||
x = Convolution2D(classes, (1, 1),
|
||||
padding='same', name='conv_preds')(x)
|
||||
x = Activation('softmax', name='act_softmax')(x)
|
||||
x = Reshape((classes,), name='reshape_2')(x)
|
||||
else:
|
||||
@@ -543,6 +557,9 @@ def _conv_block(input, filters, alpha, kernel=(3, 3), strides=(1, 1)):
|
||||
or 4D tensor with shape:
|
||||
`(samples, new_rows, new_cols, filters)` if data_format='channels_last'.
|
||||
`rows` and `cols` values might have changed due to stride.
|
||||
|
||||
# Returns
|
||||
Output tensor of block.
|
||||
"""
|
||||
channel_axis = 1 if K.image_data_format() == 'channels_first' else -1
|
||||
filters = int(filters * alpha)
|
||||
@@ -552,13 +569,11 @@ def _conv_block(input, filters, alpha, kernel=(3, 3), strides=(1, 1)):
|
||||
strides=strides,
|
||||
name='conv1')(input)
|
||||
x = BatchNormalization(axis=channel_axis, name='conv1_bn')(x)
|
||||
x = Activation(relu6, name='conv1_relu')(x)
|
||||
|
||||
return x
|
||||
return Activation(relu6, name='conv1_relu')(x)
|
||||
|
||||
|
||||
def _depthwise_conv_block(input, pointwise_conv_filters, alpha,
|
||||
depth_multiplier=1, strides=(1, 1), id=1):
|
||||
depth_multiplier=1, strides=(1, 1), block_id=1):
|
||||
"""Adds a depthwise convolution block.
|
||||
|
||||
A depthwise convolution block consists of a depthwise conv,
|
||||
@@ -588,7 +603,7 @@ def _depthwise_conv_block(input, pointwise_conv_filters, alpha,
|
||||
all spatial dimensions.
|
||||
Specifying any stride value != 1 is incompatible with specifying
|
||||
any `dilation_rate` value != 1.
|
||||
id: Integer, a unique identification designating the block number.
|
||||
block_id: Integer, a unique identification designating the block number.
|
||||
|
||||
# Input shape
|
||||
4D tensor with shape:
|
||||
@@ -602,6 +617,9 @@ def _depthwise_conv_block(input, pointwise_conv_filters, alpha,
|
||||
or 4D tensor with shape:
|
||||
`(batch, new_rows, new_cols, filters)` if data_format='channels_last'.
|
||||
`rows` and `cols` values might have changed due to stride.
|
||||
|
||||
# Returns
|
||||
Output tensor of block.
|
||||
"""
|
||||
channel_axis = 1 if K.image_data_format() == 'channels_first' else -1
|
||||
pointwise_conv_filters = int(pointwise_conv_filters * alpha)
|
||||
@@ -611,14 +629,14 @@ def _depthwise_conv_block(input, pointwise_conv_filters, alpha,
|
||||
depth_multiplier=depth_multiplier,
|
||||
strides=strides,
|
||||
use_bias=False,
|
||||
name='conv_dw_%d' % id)(input)
|
||||
x = BatchNormalization(axis=channel_axis, name='conv_dw_%d_bn' % id)(x)
|
||||
x = Activation(relu6, name='conv_dw_%d_relu' % id)(x)
|
||||
name='conv_dw_%d' % block_id)(input)
|
||||
x = BatchNormalization(axis=channel_axis, name='conv_dw_%d_bn' % block_id)(x)
|
||||
x = Activation(relu6, name='conv_dw_%d_relu' % block_id)(x)
|
||||
|
||||
x = Convolution2D(pointwise_conv_filters, (1, 1),
|
||||
padding='same',
|
||||
use_bias=False,
|
||||
strides=(1, 1),
|
||||
name='conv_pw_%d' % id)(x)
|
||||
x = BatchNormalization(axis=channel_axis, name='conv_pw_%d_bn' % id)(x)
|
||||
return Activation(relu6, name='conv_pw_%d_relu' % id)(x)
|
||||
name='conv_pw_%d' % block_id)(x)
|
||||
x = BatchNormalization(axis=channel_axis, name='conv_pw_%d_bn' % block_id)(x)
|
||||
return Activation(relu6, name='conv_pw_%d_relu' % block_id)(x)
|
||||
|
||||
Referência em uma Nova Issue
Bloquear um usuário