Python layers.conv2d方法代码示例

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def network_arg_scope(is_training=True,
                      weight_decay=cfg.train.weight_decay,
                      batch_norm_decay=0.997,
                      batch_norm_epsilon=1e-5,
                      batch_norm_scale=True):
    batch_norm_params = {
        'is_training': is_training, 'decay': batch_norm_decay,
        'epsilon': batch_norm_epsilon, 'scale': batch_norm_scale,
        'updates_collections': ops.GraphKeys.UPDATE_OPS,
        #'variables_collections': [ tf.GraphKeys.TRAINABLE_VARIABLES ],
        'trainable': cfg.train.bn_training,
    }

    with slim.arg_scope(
            [slim.conv2d, slim.separable_convolution2d],
            weights_regularizer=slim.l2_regularizer(weight_decay),
            weights_initializer=slim.variance_scaling_initializer(),
            trainable=is_training,
            activation_fn=tf.nn.relu6,
            #activation_fn=tf.nn.relu,
            normalizer_fn=slim.batch_norm,
            normalizer_params=batch_norm_params,
            padding='SAME'):
        with slim.arg_scope([slim.batch_norm], **batch_norm_params) as arg_sc:
            return arg_sc 

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def dense_block(inputs, depth, depth_bottleneck, stride, name, rate=1):
    depth_in = inputs.get_shape()[3]
    if depth == depth_in:
        if stride == 1:
            shortcut = inputs
        else:
            shortcut = layers.max_pool2d(inputs, [1, 1], stride=factor, scope=name+'_shortcut')
    else:
        shortcut = layers.conv2d(
            inputs,
            depth, [1, 1],
            stride=stride,
            activation_fn=None,
            scope=name+'_shortcut')
    if PRINT_LAYER_LOG:
        print(name+'_shortcut', shortcut.get_shape())

    residual = layers.conv2d(
        inputs, depth_bottleneck, [1, 1], stride=1, scope=name+'_conv1')
    if PRINT_LAYER_LOG:
        print(name+'_conv1', residual.get_shape())
    residual = resnet_utils.conv2d_same(
        residual, depth_bottleneck, 3, stride, rate=rate, scope=name+'_conv2')
    if PRINT_LAYER_LOG:
        print(name+'_conv2', residual.get_shape())
    residual = layers.conv2d(
        residual, depth, [1, 1], stride=1, activation_fn=None, scope=name+'_conv3')
    if PRINT_LAYER_LOG:
        print(name+'_conv3', residual.get_shape())
    output = nn_ops.relu(shortcut + residual)
    return output 

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def conv2d(inputs, c_outputs, s, name):
    output = slim.conv2d(inputs, num_outputs=c_outputs, kernel_size=[3,3], stride=s, scope=name)
    if PRINT_LAYER_LOG:
        print(name, output.get_shape())
    return output 

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def resnet_v1_backbone(inputs,
              blocks,
              is_training=True,
              output_stride=None,
              include_root_block=True,
              reuse=None,
              scope=None):
  with variable_scope.variable_scope(
      scope, 'resnet_v1', [inputs], reuse=reuse) as sc:
    end_points_collection = sc.original_name_scope + '_end_points'
    with arg_scope(
        [layers.conv2d, bottleneck, resnet_utils.stack_blocks_dense],
        outputs_collections=end_points_collection):
      with arg_scope([layers.batch_norm], is_training=is_training):
        net = inputs
        if include_root_block:
          if output_stride is not None:
            if output_stride % 4 != 0:
              raise ValueError('The output_stride needs to be a multiple of 4.')
            output_stride /= 4
          net = resnet_utils.conv2d_same(net, 64, 7, stride=2, scope='conv1')
          net = layers_lib.max_pool2d(net, [3, 3], stride=2, scope='pool1')
        net = resnet_utils.stack_blocks_dense(net, blocks, output_stride)
        # Convert end_points_collection into a dictionary of end_points.
        end_points = utils.convert_collection_to_dict(end_points_collection)

        return net, end_points 

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def inference(self, mode, inputs, scope='SenseCls'):
        is_training = mode
        with slim.arg_scope(network_arg_scope(is_training=is_training)):
            with tf.variable_scope(scope, reuse=False):
                conv0 = slim.conv2d(inputs,
                                    num_outputs=64,
                                    kernel_size=[7,7],
                                    stride=2,
                                    scope='conv0')
                if PRINT_LAYER_LOG:
                    print(conv0.name, conv0.get_shape())
                pool0 = slim.max_pool2d(conv0, kernel_size=[3, 3], stride=2, scope='pool0')
                if PRINT_LAYER_LOG:
                    print('pool0', pool0.get_shape())

                block0_0 = block(pool0, 64, 1, 'block0_0')
                block0_1 = block(block0_0, 64, 1, 'block0_1')
                block0_2 = block(block0_1, 64, 1, 'block0_2')

                block1_0 = block(block0_2, 128, 2, 'block1_0')
                block1_1 = block(block1_0, 128, 1, 'block1_1')
                block1_2 = block(block1_1, 128, 1, 'block1_2')
                block1_3 = block(block1_2, 128, 1, 'block1_3')

                block2_0 = block(block1_3, 256, 2, 'block2_0')
                block2_1 = block(block2_0, 256, 1, 'block2_1')
                block2_2 = block(block2_1, 256, 1, 'block2_2')
                block2_3 = block(block2_2, 256, 1, 'block2_3')
                block2_4 = block(block2_3, 256, 1, 'block2_4')
                block2_5 = block(block2_4, 256, 1, 'block2_5')

                block3_0 = block(block2_5, 512, 2, 'block3_0')
                block3_1 = block(block3_0, 512, 1, 'block3_1')
                block3_2 = block(block3_1, 512, 1, 'block3_2')

                net = tf.reduce_mean(block3_2, [1, 2], keepdims=True, name='global_pool_v4')
                if PRINT_LAYER_LOG:
                    print('avg_pool', net.get_shape())
                net = slim.flatten(net, scope='PreLogitsFlatten')
                net = slim.dropout(net, 0.8, is_training=is_training, scope='dropout')
                logits = fully_connected(net, cfg.classes, name='fc')
                if PRINT_LAYER_LOG:
                    print('logits', logits.get_shape())
                if is_training:
                    l2_loss = tf.add_n(tf.losses.get_regularization_losses())
                    return logits, l2_loss
                else:
                    return logits 

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def block(inputs, c_outputs, s, name):
    se_module = True
    out1 = slim.conv2d(inputs,
                       num_outputs=c_outputs,
                       kernel_size=[3,3],
                       stride=s,
                       scope=name+'_0')
    if PRINT_LAYER_LOG:
        print(name+'_0', out1.get_shape())
    output = slim.conv2d(out1,
                       num_outputs=c_outputs,
                       kernel_size=[3,3],
                       stride=1,
                       activation_fn=None,
                       scope=name+'_1')
    if PRINT_LAYER_LOG:
        print(name+'_1', output.get_shape())
    if s == 2:
        return nn_ops.relu(output)
    else:
        if se_module:
            squeeze = tf.reduce_mean(output, [1, 2], keepdims=True, name='global_pool_v4')
            if PRINT_LAYER_LOG:
                print('squeeze', squeeze.get_shape())
            fc1 = slim.conv2d(squeeze,
                            num_outputs=squeeze.get_shape()[-1] // 16,
                            normalizer_fn=None,
                            normalizer_params=None,
                            weights_regularizer=None,
                            kernel_size=[1,1],
                            stride=1,
                            activation_fn=tf.nn.relu,
                            scope=name+'_fc1')
            if PRINT_LAYER_LOG:
                print('fc1', fc1.get_shape())
            fc2 = slim.conv2d(fc1,
                            num_outputs=squeeze.get_shape()[-1],
                            normalizer_fn=None,
                            normalizer_params=None,
                            weights_regularizer=None,
                            kernel_size=[1,1],
                            stride=1,
                            activation_fn=tf.sigmoid,
                            scope=name+'_fc2')
            if PRINT_LAYER_LOG:
                print('fc2', fc2.get_shape())
            output = output * fc2
        output = nn_ops.relu(inputs + output)
        if PRINT_LAYER_LOG:
            print(name, output.get_shape())
        return output 

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def conv2d_same(inputs, num_outputs, kernel_size, stride, rate=1, scope=None):
  """Strided 2-D convolution with 'SAME' padding.

  When stride > 1, then we do explicit zero-padding, followed by conv2d with
  'VALID' padding.

  Note that

     net = conv2d_same(inputs, num_outputs, 3, stride=stride)

  is equivalent to

     net = tf.contrib.layers.conv2d(inputs, num_outputs, 3, stride=1,
     padding='SAME')
     net = subsample(net, factor=stride)

  whereas

     net = tf.contrib.layers.conv2d(inputs, num_outputs, 3, stride=stride,
     padding='SAME')

  is different when the input's height or width is even, which is why we add the
  current function. For more details, see ResnetUtilsTest.testConv2DSameEven().

  Args:
    inputs: A 4-D tensor of size [batch, height_in, width_in, channels].
    num_outputs: An integer, the number of output filters.
    kernel_size: An int with the kernel_size of the filters.
    stride: An integer, the output stride.
    rate: An integer, rate for atrous convolution.
    scope: Scope.

  Returns:
    output: A 4-D tensor of size [batch, height_out, width_out, channels] with
      the convolution output.
  """
  if stride == 1:
    return layers_lib.conv2d(
        inputs,
        num_outputs,
        kernel_size,
        stride=1,
        rate=rate,
        padding='SAME',
        scope=scope)
  else:
    kernel_size_effective = kernel_size + (kernel_size - 1) * (rate - 1)
    pad_total = kernel_size_effective - 1
    pad_beg = pad_total // 2
    pad_end = pad_total - pad_beg
    inputs = array_ops.pad(
        inputs, [[0, 0], [pad_beg, pad_end], [pad_beg, pad_end], [0, 0]])
    return layers_lib.conv2d(
        inputs,
        num_outputs,
        kernel_size,
        stride=stride,
        rate=rate,
        padding='VALID',
        scope=scope) 

# 需要导入模块: from tensorflow.contrib.layers.python.layers import layers [as 别名]
# 或者: from tensorflow.contrib.layers.python.layers.layers import conv2d [as 别名]
def bottleneck(inputs,
               depth,
               depth_bottleneck,
               stride,
               rate=1,
               outputs_collections=None,
               scope=None):
  """Bottleneck residual unit variant with BN after convolutions.

  This is the original residual unit proposed in [1]. See Fig. 1(a) of [2] for
  its definition. Note that we use here the bottleneck variant which has an
  extra bottleneck layer.

  When putting together two consecutive ResNet blocks that use this unit, one
  should use stride = 2 in the last unit of the first block.

  Args:
    inputs: A tensor of size [batch, height, width, channels].
    depth: The depth of the ResNet unit output.
    depth_bottleneck: The depth of the bottleneck layers.
    stride: The ResNet unit's stride. Determines the amount of downsampling of
      the units output compared to its input.
    rate: An integer, rate for atrous convolution.
    outputs_collections: Collection to add the ResNet unit output.
    scope: Optional variable_scope.

  Returns:
    The ResNet unit's output.
  """
  with variable_scope.variable_scope(scope, 'bottleneck_v1', [inputs]) as sc:
    depth_in = utils.last_dimension(inputs.get_shape(), min_rank=4)
    if depth == depth_in:
      shortcut = resnet_utils.subsample(inputs, stride, 'shortcut')
    else:
      shortcut = layers.conv2d(
          inputs,
          depth, [1, 1],
          stride=stride,
          activation_fn=None,
          scope='shortcut')

    residual = layers.conv2d(
        inputs, depth_bottleneck, [1, 1], stride=1, scope='conv1')
    residual = resnet_utils.conv2d_same(
        residual, depth_bottleneck, 3, stride, rate=rate, scope='conv2')
    residual = layers.conv2d(
        residual, depth, [1, 1], stride=1, activation_fn=None, scope='conv3')

    output = nn_ops.relu(shortcut + residual)

    return utils.collect_named_outputs(outputs_collections, sc.name, output)