Tensorflow问题

TypeError: 'urban' has type str, but expected one of: bytes

在前面添加"b"（例如，b'urban')，或者处理为variableName.encode("utf8")

 

Variable count_var2 already exists, disallowed. Did you mean to set reuse=True or reuse=tf.AUTO_REUSE in VarScope?

在开始的地方添加：

tf.reset_default_graph()

 

File "F:\Program Files\Python\Python36\lib\random.py", line 275, in shuffle

    x[i], x[j] = x[j], x[i]

TypeError: 'range' object does not support item assignment

后来定位原因是tfrecord是基于python2写的，里面range返回的list对象，但是再python3里面返回的是range对象。

shuffled_index = range(len(filenames)) 
random.seed(12345) 
random.shuffle(list(shuffled_index))

 

如此修改，问题解决

 

UnicodeDecodeError: 'charmap' codec can't decode byte 0x81 in position 247: character maps to <undefined>

 with open(file=filename, mode="rb") as f: 

将mode由"r"转变为"rb"

 

https://blog.csdn.net/qq_41185868/article/details/82843781

 

TypeError: 'RGB' has type str, but expected one of: bytes

 

    colorspace = b'RGB'
    channels = 3
    image_format = b'JPEG'

    example = tf.train.Example(features=tf.train.Features(feature={
        ...
        'image/colorspace': _bytes_feature(colorspace),
        ...
        'image/format': _bytes_feature(image_format),
       ...
    return example

 

 

 

 

在colorspace以及image_format中添加了'b'，问题解决。

 

TypeError: 'water' has type str, but expected one of: bytes

 

def _bytes_feature(value):
    """Wrapper for inserting bytes features into Example proto."""
    print("value值是: ", value)
    if(type(value) is not bytes):
      value = value.encode("utf8")
    return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value])) 

 

def _bytes_feature(value):
    """Wrapper for inserting bytes features into Example proto."""
    print("value值是: ", value)
    if(type(value) is not bytes):

 

 

 

注意bytes是python3中引入的一个类型；pyton2的时候后，string类型是可以当成byte来处理的，比如传输，但是到了python3时代，bytes和string是严格区分的；所有有了上面这段判断，如果不是bytes需要转换为byte（注意类型判断使用的是"is/ is not"；

 

ValueError: invalid literal for int() with base 10: 'lOOOOO'

argument = "lOOOOOO" argument = int(argument) print("complete")

后来发现那个圈其实是大写的"O"而不是"0"... ... 请问，这一条可以不写吗？后来想明白了原因，我是从PDF中直接粘出来的，PDF是一个根据图片生成的命令，会做一些莫名其妙的转换，发生了0转O的情况。