动态占位符
Tensorflow允许在占位符中具有 多个 动态（aka
None）维度。生成图形时，引擎将无法确保正确性，因此客户端负责提供正确的输入，但是它提供了很大的灵活性。
所以我要从…
x = tf.placeholder(tf.float32, shape=[None, N*M*P])y_ = tf.placeholder(tf.float32, shape=[None, N*M*P, 3])...x_image = tf.reshape(x, [-1, N, M, P, 1])
至…
# Nearly all dimensions are dynamicx_image = tf.placeholder(tf.float32, shape=[None, None, None, None, 1])label = tf.placeholder(tf.float32, shape=[None, None, 3])
由于无论如何您打算将输入重塑为5D，所以为什么不
x_image从一开始就立即使用5D 。在这一点上，的第二维
label是任意的，但是我们 保证
将与tensorflow相匹配
x_image。
去卷积中的动态形状
接下来，有趣的
tf.nn.conv3d_transpose是它的输出形状可以是动态的。所以代替这个：
# Hard-pred output shapeDeConnv1 = tf.nn.conv3d_transpose(layer1, w, output_shape=[1,32,32,7,1], ...)
… 你可以这样做：
# Dynamic output shapeDeConnv1 = tf.nn.conv3d_transpose(layer1, w, output_shape=tf.shape(x_image), ...)
这样，转置卷积可以应用于 任何 图像，并且结果将采用
x_image实际在运行时传递的形状。
请注意，静态形状
x_image为
(?, ?, ?, ?, 1)。
全卷积网络
难题的最后也是最重要的部分是使 整个网络 卷积，这也包括最后的密集层。密集层 必须 静态定义其尺寸，这会迫使整个神经网络确定输入图像的尺寸。
对我们来说幸运的是，Springenberg等人在“力求简单：全卷积网络”一文中描述了一种用CONV层替换FC层的方法。我将使用带3个
1x1x1过滤器的卷积（另请参见此问题）：
final_conv = conv3d_s1(final, weight_variable([1, 1, 1, 1, 3]))y = tf.reshape(final_conv, [-1, 3])
如果我们确保
final与
DeConnv1（和其他尺寸）具有相同的尺寸，则可以
y正确调整所需的形状：
[-1, N * M * P, 3]。
结合在一起
您的网络很大，但是所有反卷积基本上都遵循相同的模式，因此，我已将 概念验证
代码简化为一个反卷积。目的只是表明哪种网络能够处理任意大小的图像。最后说明：批次 之间的 图像尺寸可能会有所不同，但在一批内，它们必须相同。
完整代码：
sess = tf.InteractiveSession()def conv3d_dilation(tempX, tempFilter): return tf.layers.conv3d(tempX, filters=tempFilter, kernel_size=[3, 3, 1], strides=1, padding='SAME', dilation_rate=2)def conv3d(tempX, tempW): return tf.nn.conv3d(tempX, tempW, strides=[1, 2, 2, 2, 1], padding='SAME')def conv3d_s1(tempX, tempW): return tf.nn.conv3d(tempX, tempW, strides=[1, 1, 1, 1, 1], padding='SAME')def weight_variable(shape): initial = tf.truncated_normal(shape, stddev=0.1) return tf.Variable(initial)def bias_variable(shape): initial = tf.constant(0.1, shape=shape) return tf.Variable(initial)def max_pool_3x3(x): return tf.nn.max_pool3d(x, ksize=[1, 3, 3, 3, 1], strides=[1, 2, 2, 2, 1], padding='SAME')x_image = tf.placeholder(tf.float32, shape=[None, None, None, None, 1])label = tf.placeholder(tf.float32, shape=[None, None, 3])W_conv1 = weight_variable([3, 3, 1, 1, 32])h_conv1 = conv3d(x_image, W_conv1)# second convolutionW_conv2 = weight_variable([3, 3, 4, 32, 64])h_conv2 = conv3d_s1(h_conv1, W_conv2)# third convolution path 1W_conv3_A = weight_variable([1, 1, 1, 64, 64])h_conv3_A = conv3d_s1(h_conv2, W_conv3_A)# third convolution path 2W_conv3_B = weight_variable([1, 1, 1, 64, 64])h_conv3_B = conv3d_s1(h_conv2, W_conv3_B)# fourth convolution path 1W_conv4_A = weight_variable([3, 3, 1, 64, 96])h_conv4_A = conv3d_s1(h_conv3_A, W_conv4_A)# fourth convolution path 2W_conv4_B = weight_variable([1, 7, 1, 64, 64])h_conv4_B = conv3d_s1(h_conv3_B, W_conv4_B)# fifth convolution path 2W_conv5_B = weight_variable([1, 7, 1, 64, 64])h_conv5_B = conv3d_s1(h_conv4_B, W_conv5_B)# sixth convolution path 2W_conv6_B = weight_variable([3, 3, 1, 64, 96])h_conv6_B = conv3d_s1(h_conv5_B, W_conv6_B)# concatenationlayer1 = tf.concat([h_conv4_A, h_conv6_B], 4)w = tf.Variable(tf.constant(1., shape=[2, 2, 4, 1, 192]))DeConnv1 = tf.nn.conv3d_transpose(layer1, filter=w, output_shape=tf.shape(x_image), strides=[1, 2, 2, 2, 1], padding='SAME')final = DeConnv1final_conv = conv3d_s1(final, weight_variable([1, 1, 1, 1, 3]))y = tf.reshape(final_conv, [-1, 3])cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=label, logits=y))print('x_image:', x_image)print('DeConnv1:', DeConnv1)print('final_conv:', final_conv)def try_image(N, M, P, B=1): batch_x = np.random.normal(size=[B, N, M, P, 1]) batch_y = np.ones([B, N * M * P, 3]) / 3.0 deconv_val, final_conv_val, loss = sess.run([DeConnv1, final_conv, cross_entropy], feed_dict={x_image: batch_x, label: batch_y}) print(deconv_val.shape) print(final_conv.shape) print(loss) print()tf.global_variables_initializer().run()try_image(32, 32, 7)try_image(16, 16, 3)try_image(16, 16, 3, 2)
