#! /usr/bin/python
# -*- coding: utf-8 -*-
"""MobileNet for ImageNet."""
import os
import tensorflow as tf
from tensorlayer import logging
from tensorlayer.files import (assign_weights, load_npz, maybe_download_and_extract)
from tensorlayer.layers import (BatchNorm, Conv2d, DepthwiseConv2d, Flatten, GlobalMeanPool2d, Input, Reshape)
from tensorlayer.models import Model
__all__ = [
'MobileNetV1',
]
layer_names = [
'conv', 'depth1', 'depth2', 'depth3', 'depth4', 'depth5', 'depth6', 'depth7', 'depth8', 'depth9', 'depth10',
'depth11', 'depth12', 'depth13', 'globalmeanpool', 'reshape', 'out'
]
n_filters = [32, 64, 128, 128, 256, 256, 512, 512, 512, 512, 512, 512, 1024, 1024]
def conv_block(n, n_filter, filter_size=(3, 3), strides=(1, 1), name='conv_block'):
# ref: https://github.com/keras-team/keras/blob/master/keras/applications/mobilenet.py
n = Conv2d(n_filter, filter_size, strides, b_init=None, name=name + '.conv')(n)
n = BatchNorm(decay=0.99, act=tf.nn.relu6, name=name + '.batchnorm')(n)
return n
def depthwise_conv_block(n, n_filter, strides=(1, 1), name="depth_block"):
n = DepthwiseConv2d((3, 3), strides, b_init=None, name=name + '.depthwise')(n)
n = BatchNorm(decay=0.99, act=tf.nn.relu6, name=name + '.batchnorm1')(n)
n = Conv2d(n_filter, (1, 1), (1, 1), b_init=None, name=name + '.conv')(n)
n = BatchNorm(decay=0.99, act=tf.nn.relu6, name=name + '.batchnorm2')(n)
return n
def restore_params(network, path='models'):
logging.info("Restore pre-trained parameters")
maybe_download_and_extract(
'mobilenet.npz', path, 'https://github.com/tensorlayer/pretrained-models/raw/master/models/',
expected_bytes=25600116
) # ls -al
params = load_npz(name=os.path.join(path, 'mobilenet.npz'))
# for idx, net_weight in enumerate(network.all_weights):
# if 'batchnorm' in net_weight.name:
# params[idx] = params[idx].reshape(1, 1, 1, -1)
assign_weights(params[:len(network.all_weights)], network)
del params
[docs]def MobileNetV1(pretrained=False, end_with='out', name=None):
"""Pre-trained MobileNetV1 model (static mode). Input shape [?, 224, 224, 3], value range [0, 1].
Parameters
----------
pretrained : boolean
Whether to load pretrained weights. Default False.
end_with : str
The end point of the model [conv, depth1, depth2 ... depth13, globalmeanpool, out]. Default ``out`` i.e. the whole model.
name : None or str
Name for this model.
Examples
---------
Classify ImageNet classes, see `tutorial_models_mobilenetv1.py <https://github.com/tensorlayer/tensorlayer/blob/master/example/tutorial_models_mobilenetv1.py>`__
>>> # get the whole model with pretrained weights
>>> mobilenetv1 = tl.models.MobileNetV1(pretrained=True)
>>> # use for inferencing
>>> output = mobilenetv1(img1, is_train=False)
>>> prob = tf.nn.softmax(output)[0].numpy()
Extract features and Train a classifier with 100 classes
>>> # get model without the last layer
>>> cnn = tl.models.MobileNetV1(pretrained=True, end_with='reshape').as_layer()
>>> # add one more layer and build new model
>>> ni = Input([None, 224, 224, 3], name="inputs")
>>> nn = cnn(ni)
>>> nn = Conv2d(100, (1, 1), (1, 1), name='out')(nn)
>>> nn = Flatten(name='flatten')(nn)
>>> model = tl.models.Model(inputs=ni, outputs=nn)
>>> # train your own classifier (only update the last layer)
>>> train_params = model.get_layer('out').trainable_weights
Returns
-------
static MobileNetV1.
"""
ni = Input([None, 224, 224, 3], name="input")
for i in range(len(layer_names)):
if i == 0:
n = conv_block(ni, n_filters[i], strides=(2, 2), name=layer_names[i])
elif layer_names[i] in ['depth2', 'depth4', 'depth6', 'depth12']:
n = depthwise_conv_block(n, n_filters[i], strides=(2, 2), name=layer_names[i])
elif layer_names[i] == 'globalmeanpool':
n = GlobalMeanPool2d(name='globalmeanpool')(n)
elif layer_names[i] == 'reshape':
n = Reshape([-1, 1, 1, 1024], name='reshape')(n)
elif layer_names[i] == 'out':
n = Conv2d(1000, (1, 1), (1, 1), name='out')(n)
n = Flatten(name='flatten')(n)
else:
n = depthwise_conv_block(n, n_filters[i], name=layer_names[i])
if layer_names[i] == end_with:
break
network = Model(inputs=ni, outputs=n, name=name)
if pretrained:
restore_params(network)
return network