本文整理汇总了Python中tensorflow.conv3d_transpose方法的典型用法代码示例。如果您正苦于以下问题：Python tensorflow.conv3d_transpose方法的具体用法？Python tensorflow.conv3d_transpose怎么用？Python tensorflow.conv3d_transpose使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类tensorflow的用法示例。
在下文中一共展示了tensorflow.conv3d_transpose方法的1个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: model
# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import conv3d_transpose [as 别名]
def model(identities, params, is_training):
"""Model transforming embedding to voxels."""
del is_training # Unused
f_dim = params.f_dim
# Please refer to the original implementation: github.com/xcyan/nips16_PTN
# In TF replication, we use a slightly different architecture.
with slim.arg_scope(
[slim.fully_connected, conv3d_transpose],
weights_initializer=tf.truncated_normal_initializer(stddev=0.02, seed=1)):
h0 = slim.fully_connected(
identities, 4 * 4 * 4 * f_dim * 8, activation_fn=tf.nn.relu)
h1 = tf.reshape(h0, [-1, 4, 4, 4, f_dim * 8])
h1 = conv3d_transpose(
h1, f_dim * 4, [4, 4, 4], stride=2, activation_fn=tf.nn.relu)
h2 = conv3d_transpose(
h1, int(f_dim * 3 / 2), [5, 5, 5], stride=2, activation_fn=tf.nn.relu)
h3 = conv3d_transpose(
h2, 1, [6, 6, 6], stride=2, activation_fn=tf.nn.sigmoid)
return h3