最近閱讀到 Keras.backend.tensorflow_backend.py 。原由是 MobileNetV2 Keras 實現的學習，閱讀的版本是https://github.com/JonathanCMitchell/mobilenet_v2_keras 的實現，官方的另一個指定版本。
MobilenetV2 基本上是平鋪直敘，這一則是記錄對實現過程中，各個網絡層的調用細節。
-
Input
Input 起始於 from keras.layers import Input ，該方法的具體實現在 keras.engine.input_layer.py 中。- mobilenetv2.py
from keras.layers import Input - keras.layers.init.py
from ..engine import InputLayer - keras.engine.base.py
from .input_layer import Input - keras.engine.input_layer.py
def Input(*): …
詳細內容，下回分解。
- mobilenetv2.py
-
Conv2D
Conv2D 起始於 from keras.layers import Conv2D-
x = Conv2D(first_block_filters, kernel_size=3, strides=(2, 2), padding='same', use_bias=False, name='Conv1')(img_input) -
keras.layers.__init__.py
from .convolutional import * -
class Conv2D(_Conv): … class _Conv(Layer): def call(self, inputs): if self.rank == 2: outputs = K.conv2d( inputs, self.kernel, strides=self.strides, padding=self.padding, data_format=self.data_format, dilation_rate=self.dilation_rate) -
keras.backend.__init__.py
from .tensorflow_backend import * -
keras.backend.tensorflow_backend.py
def conv2d(x, kernel, strides=(1, 1), padding='valid', data_format=None, dilation_rate=(1, 1)): data_format = normalize_data_format(data_format) x, tf_data_format = _preprocess_conv2d_input(x, data_format) padding = _preprocess_padding(padding) x = tf.nn.convolution( input=x, filter=kernel, dilation_rate=dilation_rate, strides=strides, padding=padding, data_format=tf_data_format) if data_format == 'channels_first' and tf_data_format == 'NHWC': x = tf.transpose(x, (0, 3, 1, 2)) # NHWC -> NCHW return x -
tensorflow.python.ops.nn_ops.py
@tf_export("nn.convolution") def convolution( input, # pylint: disable=redefined-builtin filter, # pylint: disable=redefined-builtin padding, strides=None, dilation_rate=None, name=None, data_format=None): # pylint: disable=line-too-long """Computes sums of N-D convolutions (actually cross-correlation). This also supports either output striding via the optional `strides` parameter or atrous convolution (also known as convolution with holes or dilated convolution, based on the French word "trous" meaning holes in English) via the optional `dilation_rate` parameter. Currently, however, output striding is not supported for atrous convolutions. Specifically, in the case that `data_format` does not start with "NC", given a rank (N+2) `input` Tensor of shape [num_batches, input_spatial_shape[0], ..., input_spatial_shape[N-1], num_input_channels], a rank (N+2) `filter` Tensor of shape [spatial_filter_shape[0], ..., spatial_filter_shape[N-1], num_input_channels, num_output_channels], an optional `dilation_rate` tensor of shape [N] (defaulting to [1]*N) specifying the filter upsampling/input downsampling rate, and an optional list of N `strides` (defaulting [1]*N), this computes for each N-D spatial output position (x[0], ..., x[N-1]): ``` output[b, x[0], ..., x[N-1], k] = sum_{z[0], ..., z[N-1], q} filter[z[0], ..., z[N-1], q, k] * padded_input[b, x[0]*strides[0] + dilation_rate[0]*z[0], ..., x[N-1]*strides[N-1] + dilation_rate[N-1]*z[N-1], q] ``` where b is the index into the batch, k is the output channel number, q is the input channel number, and z is the N-D spatial offset within the filter. Here, `padded_input` is obtained by zero padding the input using an effective spatial filter shape of `(spatial_filter_shape-1) * dilation_rate + 1` and output striding `strides` as described in the @{$python/nn#Convolution$comment here}. In the case that `data_format` does start with `"NC"`, the `input` and output (but not the `filter`) are simply transposed as follows: convolution(input, data_format, **kwargs) = tf.transpose(convolution(tf.transpose(input, [0] + range(2,N+2) + [1]), **kwargs), [0, N+1] + range(1, N+1)) It is required that 1 <= N <= 3. Args: input: An N-D `Tensor` of type `T`, of shape `[batch_size] + input_spatial_shape + [in_channels]` if data_format does not start with "NC" (default), or `[batch_size, in_channels] + input_spatial_shape` if data_format starts with "NC". filter: An N-D `Tensor` with the same type as `input` and shape `spatial_filter_shape + [in_channels, out_channels]`. padding: A string, either `"VALID"` or `"SAME"`. The padding algorithm. strides: Optional. Sequence of N ints >= 1. Specifies the output stride. Defaults to [1]*N. If any value of strides is > 1, then all values of dilation_rate must be 1. dilation_rate: Optional. Sequence of N ints >= 1. Specifies the filter upsampling/input downsampling rate. In the literature, the same parameter is sometimes called `input stride` or `dilation`. The effective filter size used for the convolution will be `spatial_filter_shape + (spatial_filter_shape - 1) * (rate - 1)`, obtained by inserting (dilation_rate[i]-1) zeros between consecutive elements of the original filter in each spatial dimension i. If any value of dilation_rate is > 1, then all values of strides must be 1. name: Optional name for the returned tensor. data_format: A string or None. Specifies whether the channel dimension of the `input` and output is the last dimension (default, or if `data_format` does not start with "NC"), or the second dimension (if `data_format` starts with "NC"). For N=1, the valid values are "NWC" (default) and "NCW". For N=2, the valid values are "NHWC" (default) and "NCHW". For N=3, the valid values are "NDHWC" (default) and "NCDHW". Returns: A `Tensor` with the same type as `input` of shape `[batch_size] + output_spatial_shape + [out_channels]` if data_format is None or does not start with "NC", or `[batch_size, out_channels] + output_spatial_shape` if data_format starts with "NC", where `output_spatial_shape` depends on the value of `padding`. If padding == "SAME": output_spatial_shape[i] = ceil(input_spatial_shape[i] / strides[i]) If padding == "VALID": output_spatial_shape[i] = ceil((input_spatial_shape[i] - (spatial_filter_shape[i]-1) * dilation_rate[i]) / strides[i]). Raises: ValueError: If input/output depth does not match `filter` shape, if padding is other than `"VALID"` or `"SAME"`, or if data_format is invalid. """ # pylint: enable=line-too-long with ops.name_scope(name, "convolution", [input, filter]) as name: input = ops.convert_to_tensor(input, name="input") # pylint: disable=redefined-builtin input_shape = input.get_shape() filter = ops.convert_to_tensor(filter, name="filter") # pylint: disable=redefined-builtin filter_shape = filter.get_shape() op = Convolution( input_shape, filter_shape, padding, strides=strides, dilation_rate=dilation_rate, name=name, data_format=data_format) return op(input, filter)tensorflow.python.ops.nn_ops.py 的 def convolution() 方法的註釋，是學習convolution 原理一個很好的材料，下面這段解釋了輸入的 Tensor 如果是 [N，D_1，D_2，…，D_m-1，C]的形式，那麼filter 將會按照 [D_1，D_2，…，D_m-1，C_input，C_output]的形式生成。其中 C_output 與 filter 個數一致。
"""a rank (N+2) `input` Tensor of shape [num_batches, input_spatial_shape[0], ..., input_spatial_shape[N-1], num_input_channels], a rank (N+2) `filter` Tensor of shape [spatial_filter_shape[0], ..., spatial_filter_shape[N-1], num_input_channels, num_output_channels], """ ``` -
tensorflow.python.ops.nn_ops.py
class Convolution(object): """Helper class for convolution. Note that this class assumes that shapes of input and filter passed to __call__ are compatible with input_shape and filter_shape passed to the constructor. Arguments input_shape: static shape of input. i.e. input.get_shape(). filter_shape: static shape of the filter. i.e. filter.get_shape(). padding: see convolution. strides: see convolution. dilation_rate: see convolution. name: see convolution. data_format: see convolution. """ def __init__(self, input_shape, filter_shape, padding, strides=None, dilation_rate=None, name=None, data_format=None): """Helper function for convolution.""" # filter_shape 是 ( kernal_size, kernal_size, feature_map_channel_size，filter_number ） num_total_dims = filter_shape.ndims # input_shape 是 (NHWC）或者（NCHW） if num_total_dims is None: num_total_dims = input_shape.ndims if num_total_dims is None: raise ValueError("rank of input or filter must be known") # 空間維度，默認爲(NHWC), num_total_dims = 4 num_spatial_dims = num_total_dims - 2 try: input_shape.with_rank(num_spatial_dims + 2) except ValueError: ValueError("input tensor must have rank %d" % (num_spatial_dims + 2)) try: filter_shape.with_rank(num_spatial_dims + 2) except ValueError: ValueError("filter tensor must have rank %d" % (num_spatial_dims + 2)) if data_format is None or not data_format.startswith("NC"): input_channels_dim = input_shape[num_spatial_dims + 1] spatial_dims = range(1, num_spatial_dims + 1) else: input_channels_dim = input_shape[1] spatial_dims = range(2, num_spatial_dims + 2) if not input_channels_dim.is_compatible_with( filter_shape[num_spatial_dims]): raise ValueError( "number of input channels does not match corresponding dimension of " "filter, {} != {}".format(input_channels_dim, filter_shape[num_spatial_dims])) strides, dilation_rate = _get_strides_and_dilation_rate( num_spatial_dims, strides, dilation_rate) self.input_shape = input_shape self.filter_shape = filter_shape self.data_format = data_format self.strides = strides self.name = name self.conv_op = _WithSpaceToBatch( input_shape, dilation_rate=dilation_rate, padding=padding, build_op=self._build_op, filter_shape=filter_shape, spatial_dims=spatial_dims, data_format=data_format) def _build_op(self, _, padding): return _NonAtrousConvolution( self.input_shape, filter_shape=self.filter_shape, padding=padding, data_format=self.data_format, strides=self.strides, name=self.name) def __call__(self, inp, filter): # pylint: disable=redefined-builtin return self.conv_op(inp, filter)所以 keras.layer.Conv2D 實際用的是 tensorflow.python.ops.nn_ops.Convolution
待續
-