Question

我想在张量流r1.2中设置卷积神经网络，结构如下：

def construct_nn2(input_layer, net2={}):

    input_layer1 = tf.nn.dilation2d(
        input=input_layer,
        strides=[1, 1, 1, 1],
        rates=[1, 1, 1, 1],
        filter=[1.0, 1.0, 1],
        padding='SAME',
        name='dil'
    )

    net2['conv11'] = tf.layers.conv2d(
        inputs=input_layer1,
        filters=64,
        kernel_size=[5, 5],
        padding='same',
        activation=tf.tanh,
        name='conv11'
    )

    net2['conv12'] = tf.layers.conv2d(
        inputs=net2['conv11'],
        filters=64,
        kernel_size=[3, 3],
        padding='same',
        activation=tf.tanh,
        name='conv12'
    )

    net2['logits']= tf.layers.conv2d(
        inputs=net2['conv12'],
        filters=1,
        kernel_size=[3, 3],
        padding='same',
        activation=tf.sigmoid,
        name='logits'
    )

return net2['logits']

扩张层吐出这些错误：

Traceback (most recent call last):
  File "/home/test/Dropbox/occlusion_thesis/occ_small _2_add/main.py", line 137, in <module>
    tn_prediction = construct_nn2(t_img)
  File "/home/test/Dropbox/occlusion_thesis/occ_small _2_add/main.py", line 18, in construct_nn2
    name='dil'
  File "/home/test/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/ops/gen_nn_ops.py", line 860, in dilation2d
    name=name)
  File "/home/test/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op
    op_def=op_def)
  File "/home/test/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2508, in create_op
    set_shapes_for_outputs(ret)
  File "/home/test/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1873, in set_shapes_for_outputs
    shapes = shape_func(op)
  File "/home/test/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1823, in call_with_requiring
    return call_cpp_shape_fn(op, require_shape_fn=True)
  File "/home/test/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/common_shapes.py", line 610, in call_cpp_shape_fn
    debug_python_shape_fn, require_shape_fn)
  File "/home/test/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/common_shapes.py", line 676, in _call_cpp_shape_fn_impl
    raise ValueError(err.message)
ValueError: Shape must be rank 3 but is rank 1 for 'dil' (op: 'Dilation2D') with input shapes: [1,248,360,128], [3].

我必须改变错误不会引起的。我已经阅读了dilation2d，但我仍然没有任何线索，我需要改变的是，这个层可以至少没有错误地工作。

编辑：

过滤器必须是具有相同形状的张量。

input_layer1 = tf.nn.dilation2d(
    input=input_layer,
    strides=[1, 1, 1, 1],
    rates=[1, 1, 1, 1],
    filter=tf.squeeze(input_layer, axis=0),
    padding='SAME',
    name='dil'
)

重构过滤器确实至少要运行。

Tensorflow功能：dilation2d（）引发形状误差

0 个答案: