Question

我尝试写一个tfrecords并阅读它，我已经获得了'train.tfrecords'文件，但是当我使用这个函数来阅读它时

image_size=224
def read_and_decode(filename,batch_size):
  filename_queue=tf.train.string_input_producer([filename])
  reader=tf.TFRecordReader()
  _,serialized_example=reader.read(filename_queue)#返回文件名和文件
  features=tf.parse_single_example(serialized_example,features={
    "label": tf.FixedLenFeature([],tf.int64),
    "img_raw":tf.FixedLenFeature([],tf.string),
  })
  img=tf.decode_raw(features['img_raw'],tf.uint8)
  img=tf.cast(img,tf.float32)
  img=tf.reshape(img,[image_size,image_size,3])
  img = tf.random_crop(img, [image_size, image_size, 3])
  img = tf.image.random_flip_left_right(img)
  img=tf.image.per_image_standardization(img)
  label=tf.cast(features['label'],tf.int32)
  img_batch, label_batch = tf.train.shuffle_batch([img, label],
                                                batch_size=batch_size,    num_threads=10, capacity=16 * batch_size,
                                                 min_after_dequeue=8*batch_size)
  label_batch= tf.reshape(label_batch, [batch_size, 1])
  indices = tf.reshape(tf.range(0, batch_size, 1), [batch_size, 1])
  label_batch = tf.sparse_to_dense(
    tf.concat(values=[indices, label_batch], axis=1),
    [batch_size, 3], 1.0, 0.0)
  assert len(img_batch.get_shape()) == 4
  assert img_batch.get_shape()[0] == batch_size
  assert img_batch.get_shape()[-1] == 3
  assert len(label_batch.get_shape()) == 2
  assert label_batch.get_shape()[0] == batch_size
  assert label_batch.get_shape()[1] == 3

  # Display the training images in the visualizer.
  tf.summary.image('images', img_batch)
  return img_batch, label_batch

错误是：

Caused by op 'Reshape', defined at:
File "C:/Users/Administrator/Desktop/tensorflow/ResNet/main.py", line  176, in <module>tf.app.run()
File"C:\ProgramFiles\Python35\lib\sitepackages\tensorflow\python\platform\app.py", line 48, in run
_sys.exit(main(_sys.argv[:1] + flags_passthrough))
 File "C:/Users/Administrator/Desktop/tensorflow/ResNet/main.py", line 169, in main
train(hps)
File "C:/Users/Administrator/Desktop/tensorflow/ResNet/main.py", line 31, in train
images, labels = read_and_decode('train.tfrecords', hps.batch_size)
File "C:\Users\Administrator\Desktop\tensorflow\ResNet\input.py", line 15, in read_and_decode
img=tf.reshape(img,[image_size,image_size,3])
 File "C:\Program Files\Python35\lib\site-packages\tensorflow\python\ops\gen_array_ops.py", line 2510, in reshape
name=name)
  File "C:\Program Files\Python35\lib\site-packages\tensorflow\python\framework\op_def_library.py", line 768, in apply_op
op_def=op_def)
File "C:\Program Files\Python35\lib\site-packages\tensorflow\python\framework\ops.py", line 2336, in create_op
original_op=self._default_original_op, op_def=op_def)
File "C:\Program Files\Python35\lib\site-packages\tensorflow\python\framework\ops.py", line 1228, in __init__
self._traceback = _extract_stack()
InvalidArgumentError (see above for traceback): Input to reshape is a tensor with 154587 values, but the requested shape has 150528
 [[Node: Reshape = Reshape[T=DT_FLOAT, Tshape=DT_INT32, _device="/job:localhost/replica:0/task:0/cpu:0"](Cast, Reshape/shape)]]

我在互联网上搜索了一段时间，我认为问题发生在这些代码上，但我不知道如何修复它，请帮帮我，谢谢

img=tf.decode_raw(features['img_raw'],tf.uint8)
img=tf.cast(img,tf.float32)
img=tf.reshape(img,[image_size,image_size,3])

Answer 1

您可以在日志中清楚地看到这一点，正如@Mathias Rav在评论中所解释的那样。

paths: {
      // paths serve as alias
      'npm:': 'https://unpkg.com/'
    },

TFrecords的大小是154587：227x227x3

但是，您提供的张量形状的大小是150528：224x224x3

首先检查模型所需的输入尺寸，然后相应地调整图像。或者，

使用tf.image.resize_images（）或
再次准备TFRecords，大小为224x224x3或
如果型号输入尺寸为227x227x3，则只需将image_size更改为227.

我希望这会有所帮助。

InvalidArgumentError（参见上面的回溯）：重塑的输入是一个具有154587值的张量，但请求的形状有150528

1 个答案: