tf.nn.conv2d_transpose 给出 InvalidArgumentError: Conv2DCustomBackpropInput: input and filter must have the same depth答案

【问题标题】：tf.nn.conv2d_transpose gives InvalidArgumentError: Conv2DCustomBackpropInput: input and filter must have the same depthtf.nn.conv2d_transpose 给出 InvalidArgumentError: Conv2DCustomBackpropInput: input and filter must have the same depth
【发布时间】：2018-05-19 15:04:33
【问题描述】：

我遇到了让tf.nn.conv2d_transpose 正常工作的问题。这是我正在尝试做的一个小复制：

import tensorflow as tf
import numpy as np

# Shape (2, 3, 3, 1) == (batch_sz, height, width, channels)
inp = tf.Variable(np.array(
    [
        [
            [[1], [2], [3]],
            [[2], [3], [4]],
            [[7], [8], [9]]
        ],
        [
            [[3], [2], [1]],
            [[2], [7], [2]],
            [[3], [2], [0]]
        ]
    ], dtype = np.float32
))
# Shape (5, 5, 3, 1) == (kH, kW, out_channels, in_channels)
ker = tf.Variable(np.array(
    [
        [[[1],[2],[1]], [[2],[2],[2]], [[1],[2],[1]], [[2],[1],[1]], [[1],[1],[1]]],
        [[[1],[2],[1]], [[2],[2],[2]], [[1],[2],[1]], [[2],[1],[1]], [[1],[1],[1]]],
        [[[1],[2],[1]], [[2],[2],[2]], [[1],[2],[1]], [[2],[1],[1]], [[1],[1],[1]]],
        [[[1],[2],[1]], [[2],[2],[2]], [[1],[2],[1]], [[2],[1],[1]], [[1],[1],[1]]],
        [[[1],[2],[1]], [[2],[2],[2]], [[1],[2],[1]], [[2],[1],[1]], [[1],[1],[1]]]
    ], dtype = np.float32
))
out = tf.nn.conv2d_transpose(inp, ker, (2, 7, 7, 1), (1, 1, 1, 1), padding='SAME', data_format='NHWC', name='conv_transpose')

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    output, kernel, input = sess.run([out, ker, inp])

我想要的是使用三个 5x5x1 过滤器对 3x3x1 输入执行转置卷积。我希望输出的形状为 7x7x3 - 但相反，我收到一条错误消息：

InvalidArgumentError: Conv2DCustomBackpropInput: input and filter must have the same depth
     [[Node: conv_transpose_2 = Conv2DBackpropInput[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="SAME", strides=[1, 1, 1, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/device:CPU:0"](conv_transpose_2/output_shape, Variable_21/read, Variable_20/read)]]

输入和过滤器深度不是都等于1吗？我看不出我做错了什么 - 任何提示都会非常感激。我特别想使用tf.nn.conv2d_transpose 而不是tf.layers.conv2d_transpose。

【问题讨论】：

你扩展了数组的第 0 维和第 3 维吗？
抱歉听起来像个菜鸟。不我没有。为什么我需要这样做？我的数组是否还不是 conv2d_transpose 所需的格式 - [batches, in_height, in_width, channels] 用于输入，[filter_height, filter_width, out_channels, in_channels] 用于过滤器 -> 文档：tensorflow.org/api_docs/python/tf/nn/conv2d_transpose

标签： python tensorflow conv-neural-network convolution deconvolution

【解决方案1】：

此问题与Stack Overflow Issue 类似。

您应该对代码进行以下更改才能运行。

inp 的形状应该是 (2, 3, 3, 3) 而不是 (2, 3, 3, 1)
ker 的形状应该是 (5,5,1,3) 而不是 (5,5,3,1)
只有在Output Shape 与Input Shape 不同时，填充应设置为“VALID”而不是“SAME”。

下面提到的是工作代码（为了便于实施，用零做了）：

import tensorflow as tf
import numpy as np

# Shape (2, 3, 3, 3) == (batch_sz, height, width, channels)
inp = tf.Variable(np.array(np.zeros((2, 3, 3, 3)) , dtype = np.float32))

# Shape (5, 5, 3, 1) == (kH, kW, out_channels, in_channels)
ker = tf.Variable(np.zeros((5,5,1,3)) , dtype = np.float32)

out = tf.nn.conv2d_transpose(value = inp, filter = ker, output_shape=(2, 7, 7, 1), 
                             strides=(1, 1, 1, 1), padding='VALID', data_format='NHWC', name='conv_transpose')

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    output, kernel, input = sess.run([out, ker, inp])

【讨论】：