我在尝试使用TensorFlow中的常规神经网络读取png格式的图像时遇到了问题。我在读取数据步骤中无法处理标签，因此出现了错误。

我首先认为问题出在我创建批处理队列和洗牌时。这在tf.train.shuffle_batch中出现了问题，它期望的是1-D的输入，但实际上得到了3-D的形状。

我修复了这个问题，但随后在损失函数中又出现了另一个错误

子代码在这里：链接

问题是：

Traceback (most recent call last): File “SVHN.py”, line 280, in tf.app.run() File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/platform/app.py”, line 48, in run _sys.exit(main(_sys.argv[:1] + flags_passthrough)) File “SVHN.py”, line 269, in main train_op = SVHN_architecture_AND_optimize() File “SVHN.py”, line 203, in SVHN_architecture_AND_optimize cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits = logits, labels = labels) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/nn_ops.py”, line 1742, in sparse_softmax_cross_entropy_with_logits precise_logits, labels, name=name) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/gen_nn_ops.py”, line 2418, in _sparse_softmax_cross_entropy_with_logits features=features, labels=labels, name=name) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py”, line 768, in apply_op op_def=op_def) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/ops.py”, line 2338, in create_op set_shapes_for_outputs(ret) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/ops.py”, line 1719, in set_shapes_for_outputs shapes = shape_func(op) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/ops.py”, line 1669, in call_with_requiring return call_cpp_shape_fn(op, require_shape_fn=True) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/common_shapes.py”, line 610, in call_cpp_shape_fn debug_python_shape_fn, require_shape_fn) File “/home/ashwaq/anaconda3/lib/python3.6/site-packages/tensorflow/python/framework/common_shapes.py”, line 676, in _call_cpp_shape_fn_impl raise ValueError(err.message) ValueError: Dimensions must be equal, but are 1 and 128 for ‘SparseSoftmaxCrossEntropyWithLogits/SparseSoftmaxCrossEntropyWithLogits’ (op: ‘SparseSoftmaxCrossEntropyWithLogits’) with input shapes: [1,10], [128].

回答：

错误的原因是只有1张输入图像被传递给了SparseSoftmaxEntropyWithLogits，而不是一个完整的批次。导致问题的是这行代码：

# 将单个输入的`image`更改为批处理的`images`images = tf.reshape(images, [-1, IMAGE_SIZE, IMAGE_SIZE, depth])