GroupConv implement

[Windaway]

class GroupConv(Layer):
    def __init__(
        self,
        layer = None,
        filters=0,
        size=3,
        stride=2,
        groups=2,
        act = tf.identity,
        name ='groupconv',
    ):
        Layer.__init__(self, name=name)
        self.inputs = layer.outputs
        groupConv = lambda i, k: tf.nn.conv2d(i, k, strides=[1, stride, stride, 1], padding='VALID')
        channels = int(self.inputs.get_shape()[-1])
        with tf.variable_scope(name) as vs:
            We = tf.get_variable(name='weights', shape=[size, size, channels / groups, filters],
                                       initializer=tf.truncated_normal_initializer(stddev=0.03),
                                       dtype=tf.float32, trainable=True)
            bi = tf.get_variable(name='biases', shape=[filters, ],
                                      initializer=tf.constant_initializer(value=0.1),
                                      dtype=tf.float32, trainable=True)
        if groups == 1:
            conv = groupConv(self.inputs, We)
        else:
            inputGroups = tf.split(axis=3, num_or_size_splits=groups, value=self.inputs)
            weightsGroups = tf.split(axis=3, num_or_size_splits=groups, value=We)
            convGroups = [groupConv(i, k) for i, k in zip(inputGroups, weightsGroups)]
            conv = tf.concat(axis=3, values=convGroups)
        if name is not None:
            conv = tf.add(conv, bi, name=name)
        else:
            conv = tf.add(conv, bi)
        self.outputs =act(conv)
        self.all_layers = list(layer.all_layers)
        self.all_params = list(layer.all_params)
        self.all_drop = dict(layer.all_drop)
        self.all_layers.extend( [self.outputs] )
        self.all_params.extend( [We, bi] )

Caffe中分组卷积的实现，一个减少过拟合的卷积设计。之前孪生网络论文实现用过这个。

[zsdonghao]

@Windaway hi, 我改了一下代码，并加了一些描述，你帮忙看看对不对？

class GroupConv2d(Layer):
    """The :class:`GroupConv2d` class is 2D grouped convolution, see `here <https://blog.yani.io/filter-group-tutorial/>`__.

    Parameters
    --------------
    layer : :class:`Layer`
        Previous layer.
    n_filter : int
        The number of filters.
    filter_size : int
        The filter size.
    stride : int
        The stride step.
    n_group : int
        The number of groups.
    act : activation function
        The activation function of this layer.
    padding : str
        The padding algorithm type: "SAME" or "VALID".
    W_init : initializer
        The initializer for the weight matrix.
    b_init : initializer or None
        The initializer for the bias vector. If None, skip biases.
    W_init_args : dictionary
        The arguments for the weight matrix initializer.
    b_init_args : dictionary
        The arguments for the bias vector initializer.
    name : str
        A unique layer name.
    """
    def __init__(
        self,
        layer = None,
        n_filter =0,
        filter_size=(3, 3),
        strides=(2, 2),
        n_group=2,
        act = tf.identity,
        padding='SAME',
        W_init=tf.truncated_normal_initializer(stddev=0.02),
        b_init=tf.constant_initializer(value=0.0),
        W_init_args=None,
        b_init_args=None,
        name ='groupconv',
    ): # Windaway 
        Layer.__init__(self, name=name)
        self.inputs = layer.outputs
        groupConv = lambda i, k: tf.nn.conv2d(i, k, strides=[1, strides[0], strides[1], 1], padding= padding)
        channels = int(self.inputs.get_shape()[-1])
        with tf.variable_scope(name) as vs:
            We = tf.get_variable(name='weights', shape=[filter_size[0], filter_size[1], channels / n_group, n_filter],  initializer=W_init, dtype=D_TYPE,  trainable=True, **W_init_args)
            if bi:
                bi = tf.get_variable(name='biases', shape=[n_filter, ],
                                      initializer=tf.constant_initializer(value=0.1),
                                      initializer=b_init, dtype=D_TYPE, trainable=True, **b_init_args )
        if groups == 1:
            conv = groupConv(self.inputs, We)
        else:
            inputGroups = tf.split(axis=3, num_or_size_splits=n_group, value=self.inputs)
            weightsGroups = tf.split(axis=3, num_or_size_splits=n_group, value=We)
            convGroups = [groupConv(i, k) for i, k in zip(inputGroups, weightsGroups)]
            conv = tf.concat(axis=3, values=convGroups)
        if b_init is not None:
            conv = tf.add(conv, bi, name='add')

        self.outputs =act(conv)
        self.all_layers = list(layer.all_layers)
        self.all_params = list(layer.all_params)
        self.all_drop = dict(layer.all_drop)
        self.all_layers.extend( [self.outputs] )
        if b_init is not None:
            self.all_params.extend( [We, bi] )
        else:
            self.all_params.extend( [We] )

[zsdonghao]

Final Version:

class GroupConv2d(Layer):
    """The :class:`GroupConv2d` class is 2D grouped convolution, see `here <https://blog.yani.io/filter-group-tutorial/>`__.

    Parameters
    --------------
    layer : :class:`Layer`
        Previous layer.
    n_filter : int
        The number of filters.
    filter_size : int
        The filter size.
    stride : int
        The stride step.
    n_group : int
        The number of groups.
    act : activation function
        The activation function of this layer.
    padding : str
        The padding algorithm type: "SAME" or "VALID".
    W_init : initializer
        The initializer for the weight matrix.
    b_init : initializer or None
        The initializer for the bias vector. If None, skip biases.
    W_init_args : dictionary
        The arguments for the weight matrix initializer.
    b_init_args : dictionary
        The arguments for the bias vector initializer.
    name : str
        A unique layer name.
    """

    def __init__(
            self,
            layer=None,
            n_filter=32,
            filter_size=(3, 3),
            strides=(2, 2),
            n_group=2,
            act=tf.identity,
            padding='SAME',
            W_init=tf.truncated_normal_initializer(stddev=0.02),
            b_init=tf.constant_initializer(value=0.0),
            W_init_args=None,
            b_init_args=None,
            name='groupconv',
    ):  # Windaway
        if W_init_args is None:
            W_init_args = {}
        if b_init_args is None:
            b_init_args = {}

        Layer.__init__(self, name=name)
        self.inputs = layer.outputs
        groupConv = lambda i, k: tf.nn.conv2d(i, k, strides=[1, strides[0], strides[1], 1], padding=padding)
        channels = int(self.inputs.get_shape()[-1])
        with tf.variable_scope(name) as vs:
            We = tf.get_variable(
                name='weights',
                shape=[filter_size[0], filter_size[1], channels / n_group, n_filter],
                initializer=W_init,
                dtype=D_TYPE,
                trainable=True,
                **W_init_args)
            if b_init:
                bi = tf.get_variable(
                    name='biases', shape=[
                        n_filter,
                    ], initializer=b_init, dtype=D_TYPE, trainable=True, **b_init_args)
        if n_group == 1:
            conv = groupConv(self.inputs, We)
        else:
            inputGroups = tf.split(axis=3, num_or_size_splits=n_group, value=self.inputs)
            weightsGroups = tf.split(axis=3, num_or_size_splits=n_group, value=We)
            convGroups = [groupConv(i, k) for i, k in zip(inputGroups, weightsGroups)]
            conv = tf.concat(axis=3, values=convGroups)
        if b_init is not None:
            conv = tf.add(conv, bi, name='add')

        self.outputs = act(conv)
        self.all_layers = list(layer.all_layers)
        self.all_params = list(layer.all_params)
        self.all_drop = dict(layer.all_drop)
        self.all_layers.extend([self.outputs])
        if b_init is not None:
            self.all_params.extend([We, bi])
        else:
            self.all_params.extend([We])