??tf.nn.separable_conv2d

來自專欄 TensorFlow

Args:

• input: 4-D Tensor with shape [batch, in_height, in_width, in_channels].
• depthwise_filter: 4-D Tensor with shape [filter_height, filter_width, in_channels, channel_multiplier]. Contains in_channels convolutional filters of depth 1.
• pointwise_filter: 4-D Tensor with shape [1, 1, channel_multiplier * in_channels, out_channels]. Pointwise filter to mix channels after depthwise_filter has convolved spatially.
• strides: 1-D of size 4. The strides for the depthwise convolution for each dimension of input.
• padding: A string, either VALID or SAME. The padding algorithm.
• name: A name for this operation (optional).

import tensorflow as tftest API: tf.nn.separable_conv2da = tf.Variable(tf.truncated_normal([3, 3, 3, 2]))b = tf.Variable(tf.truncated_normal([1, 1, 6, 8]))input = tf.placeholder(tf.float32, shape = [2, 5, 5, 3])output = tf.nn.separable_conv2d(input, #4-D Tensor with shape according to data_format a, #4-D Tensor with shape [filter_height, filter_width, in_channels, channel_multiplier]. Contains in_channels convolutional filters of depth 1 b, #4-D Tensor with shape [1, 1, channel_multiplier * in_channels, out_channels]. Pointwise filter to mix channels after depthwise_filter has convolved spatially strides = [1,1,1,1], padding = SAME)print(output.get_shape().as_list())

推薦閱讀：