tensorflow实现 Spatial Pyramid Poo

实现的代码如下：

# Spatial Pyramid Pooling block

# https://arxiv.org/abs/1406.4729

def spatial_pyramid_pool(previous_conv, num_sample, previous_conv_size, out_pool_size):

  """

  previous_conv: a tensor vector of previous convolution layer

  num_sample: an int number of image in the batch

  previous_conv_size: an int vector [height, width] of the matrix features size of previous convolution layer

  out_pool_size: a int vector of expected output size of max pooling layer

  returns: a tensor vector with shape [1 x n] is the concentration of multi-level pooling

  """

  for i in range(len(out_pool_size)):

    h_strd = h_size = math.ceil(float(previous_conv_size[0]) / out_pool_size[i])

    w_strd = w_size = math.ceil(float(previous_conv_size[1]) / out_pool_size[i])

    pad_h = int(out_pool_size[i] * h_size - previous_conv_size[0])

    pad_w = int(out_pool_size[i] * w_size - previous_conv_size[1])

    new_previous_conv = tf.pad(previous_conv, tf.constant([[0, 0], [0, pad_h], [0, pad_w], [0, 0]]))

    max_pool = tf.nn.max_pool(new_previous_conv,

                  ksize=[1,h_size, h_size, 1],

                  strides=[1,h_strd, w_strd,1],

                  padding='SAME')

    if (i == 0):

      spp = tf.reshape(max_pool, [num_sample, -1])

    else:

      spp = tf.concat(axis=1, values=[spp, tf.reshape(max_pool, [num_sample, -1])])

  return spp