将张量保存为JPEG图像--Python / TensorFlow

时间:2017-10-02 13:55:33

标签: python python-2.7 image-processing tensorflow tensor

我有一个函数可以返回一个名为图层的变量 - 图像格式为:

<tf.Tensor 'Conv2D_1:0' shape=(?, 16, 16, 1) dtype=float32>

我需要将这些图像保存在.jpeg中。

到目前为止,我一直想到这样做:

# Reshape into tf.image.encode_jpeg format
images = tf.reshape(tf.cast(layer, tf.uint8), [16, 16, 1])

# Encode
images_encode = tf.image.encode_jpeg(images)

# Create a files name
fname = tf.constant('datetime.now() + ".jpeg"')

# Write files
fwrite = tf.write_file(fname, images_encode)

train_batch_size = 300

在session = tf.Session()

# That means it will only scroll through my 300 images...
x_batch, y_true_batch = next_batch_size(train_batch_size)

feed_dict_train = {x: x_batch, y_true: y_true_batch}

result = session.run(fwrite, feed_dict=feed_dict_train)

但我收到以下错误:

InvalidArgumentError (see above for traceback): Input to reshape is a tensor with 76800 values, but the requested shape has 256
    [[Node: Reshape_7 = Reshape[T=DT_UINT8, Tshape=DT_INT32, _device="/job:localhost/replica:0/task:0/cpu:0"](Cast_7, Reshape_7/shape)]]

我的占位符是:

# Placeholder variable for the input images
x = tf.placeholder(tf.float32, shape=[None, img_size_flat], name='x')

# Reshape 'x'
x_image = tf.reshape(x, [-1, img_size, img_size, num_channels])

# Placeholder variable for the true labels associated with the images
y_true = tf.placeholder(tf.float32, shape=[None, num_classes], name='y_true')

知道如何解决这个问题,或者我可以应用哪些方法来保存图像?

1 个答案:

答案 0 :(得分:1)

您使用的批量大小为76800/256 = 300

重塑功能正在尝试将整批重塑为(16,16,1)张量:这是不可能的。

如果您想保存单张图片,您有两种选择:

  1. 提取批次tf.reshape(tf.cast(layer[0], tf.uint8), [16, 16, 1])
  2. 的第一个元素
  3. 将批量大小设置为1
  4. 如果您要保存整批图像,则必须循环批处理(使用tf.map_fn)并单独编码图像(因为tf.image.encode_jpeg适用于单个图像)。然后,从python中提取每个编码图像并将其保存到磁盘。