&引用；TypeError:不允许将'tf.Tensor'用作Python'bool'；在数据集上调用map函数时

我试图加载和处理图像，并从每个图像中学习一个独特的裁剪因子。我不断收到一个错误，说明我不能将张量用作Python布尔值

对于每个图像，我想从图像中心设置一行像素的阈值，并计算超过某个阈值的像素百分比。我想用这个百分比作为作物系数

def preprocess_image(image):
    image = tf.image.decode_png(image, channels=3)
    print(tf.shape(image))
    halfpix = tf.shape(image)[0]//2
    row = tf.cast(tf.math.greater(image[:, :, 0][halfpix, :], 3), tf.float32)
    hor_scale_factor = tf.math.reduce_mean(row)
    image = tf.image.central_crop(image, hor_scale_factor)

    return image

def load_and_preprocess_image(path):
    image = tf.io.read_file(path)
    return preprocess_image(image)

train_image_ds = train_path_ds.map(load_and_preprocess_image, num_parallel_calls=AUTOTUNE)

我希望没有错误。我收到：“类型错误：不允许将

tf.Tensor

用作Python

bool

。如果t不是None，则使用

而不是
if t:
来测试是否定义了张量，并使用诸如tf.cond的TensorFlow操作来执行以张量值为条件的子图。”

完整跟踪：

TypeError                                 Traceback (most recent call last) <ipython-input-89-b7d7da47ff6e> in <module>
----> 1 train_image_ds = train_path_ds.map(load_and_preprocess_image, num_parallel_calls=AUTOTUNE)
      2 test_image_ds = test_path_ds.map(load_and_preprocess_image, num_parallel_calls=AUTOTUNE)

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\data\ops\dataset_ops.py in map(self, map_func, num_parallel_calls)    1144     else:    1145   return ParallelMapDataset(
-> 1146           self, map_func, num_parallel_calls, preserve_cardinality=True)    1147     1148   def flat_map(self, map_func):

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\data\ops\dataset_ops.py in __init__(self, input_dataset, map_func, num_parallel_calls, use_inter_op_parallelism, preserve_cardinality, use_legacy_function)   3262         self._transformation_name(),    3263         dataset=input_dataset,
-> 3264         use_legacy_function=use_legacy_function)    3265     self._num_parallel_calls = ops.convert_to_tensor(    3266         num_parallel_calls, dtype=dtypes.int32, name="num_parallel_calls")

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\data\ops\dataset_ops.py in __init__(self, func, transformation_name, dataset, input_classes, input_shapes, input_types, input_structure, add_to_graph, use_legacy_function, defun_kwargs)    2589       resource_tracker = tracking.ResourceTracker()    2590       with tracking.resource_tracker_scope(resource_tracker):
-> 2591         self._function = wrapper_fn._get_concrete_function_internal()    2592         if add_to_graph:    2593           self._function.add_to_graph(ops.get_default_graph())

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\eager\function.py in _get_concrete_function_internal(self, *args, **kwargs)    1364     """Bypasses error checking when getting a graph function."""    1365   graph_function = self._get_concrete_function_internal_garbage_collected(
-> 1366         *args, **kwargs)    1367     # We're returning this concrete function to someone, and they may keep a    1368     # reference to the FuncGraph without keeping a reference to the

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\eager\function.py in _get_concrete_function_internal_garbage_collected(self, *args,
**kwargs)    1358     if self.input_signature:    1359       args, kwargs = None, None
-> 1360     graph_function, _, _ = self._maybe_define_function(args, kwargs)    1361     return graph_function    1362 

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\eager\function.py in _maybe_define_function(self, args, kwargs)    1646       graph_function = self._function_cache.primary.get(cache_key, None)    1647       if graph_function is None:
-> 1648         graph_function = self._create_graph_function(args, kwargs)    1649         self._function_cache.primary[cache_key] = graph_function    1650       return graph_function, args, kwargs

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\eager\function.py in _create_graph_function(self, args, kwargs, override_flat_arg_shapes)    1539             arg_names=arg_names,    1540             override_flat_arg_shapes=override_flat_arg_shapes,
-> 1541             capture_by_value=self._capture_by_value),    1542         self._function_attributes)    1543 

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\framework\func_graph.py in func_graph_from_py_func(name, python_func, args, kwargs, signature, func_graph, autograph, autograph_options, add_control_dependencies, arg_names, op_return_value, collections, capture_by_value, override_flat_arg_shapes)
    714                                           converted_func)
    715 
--> 716       func_outputs = python_func(*func_args, **func_kwargs)
    717 
    718       # invariant: `func_outputs` contains only Tensors, CompositeTensors,

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\data\ops\dataset_ops.py in wrapper_fn(*args)    2583           attributes=defun_kwargs)    2584       def wrapper_fn(*args):  # pylint: disable=missing-docstring
-> 2585         ret = _wrapper_helper(*args)    2586         ret = self._output_structure._to_tensor_list(ret)    2587         return [ops.convert_to_tensor(t) for t in ret]

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\data\ops\dataset_ops.py in _wrapper_helper(*args)    2528         nested_args = (nested_args,) 2529 
-> 2530       ret = func(*nested_args)    2531       # If `func` returns a list of tensors, `nest.flatten()` and    2532       # `ops.convert_to_tensor()` would conspire to attempt to stack

<ipython-input-86-61d7ec60892c> in load_and_preprocess_image(path)
      1 def load_and_preprocess_image(path):
      2     image = tf.io.read_file(path)
----> 3     return preprocess_image(image)

<ipython-input-85-4f3b9475e191> in preprocess_image(image)
     11     hor_scale_factor = tf.math.reduce_mean(row)
     12 #     print(hor_scale_factor)
---> 13     image = tf.image.central_crop(image, hor_scale_factor)
     14 #     print(type(hor_scale_factor))
     15     image = tf.image.resize(image, target_im_size, preserve_aspect_ratio=True) # Resize to final dimensions

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\ops\image_ops_impl.py in central_crop(image, central_fraction)
    641   with ops.name_scope(None, 'central_crop', [image]):
    642     image = ops.convert_to_tensor(image, name='image')
--> 643     if central_fraction <= 0.0 or central_fraction > 1.0:
    644       raise ValueError('central_fraction must be within (0, 1]')
    645     if central_fraction == 1.0:

c:\users\toby-pc\documents\code\blindness_kaggle\my_env\lib\site-packages\tensorflow\python\framework\ops.py in __bool__(self)
    696       `TypeError`.
    697     """
--> 698     raise TypeError("Using a `tf.Tensor` as a Python `bool` is not allowed. "
    699                     "Use `if t is not None:` instead of `if t:` to test if a "
    700                     "tensor is defined, and use TensorFlow ops such as "

TypeError: Using a `tf.Tensor` as a Python `bool` is not allowed. Use `if t is not None:` instead of `if t:` to test if a tensor is defined, and use TensorFlow ops such as tf.cond to execute subgraphs conditioned on the value of a tensor.

TypeError回溯（最近一次调用最后一次）
---->1列车图像\u ds=列车路径\u ds.map（加载和预处理图像，并行调用数=自动调谐）
2 test\u image\u ds=test\u path\u ds.map（加载和预处理\u image，num\u parallel\u calls=AUTOTUNE）
c:\users\toby pc\documents\code\blindness\u kaggle\my_env\lib\site packages\tensorflow\python\data\ops\dataset\ops.py in map（self，map\u func，num\u parallel\u calls）1144其他：1145返回ParallelMapDataset(
->1146 self，map_func，num_parallel_调用，preserve_cardinality=True）1147 1148 def flat_map（self，map_func）：
c:\users\toby pc\documents\code\blindness\u kaggle\my_env\lib\site packages\tensorflow\python\data\ops\dataset\ops.py in uuuu init_uuuuuuuu（self，input_dataset，map_func，num_parallel_u调用，使用u inter o o o o o并行，保留基数，使用o遗留o函数）3262; 3262; 3262; self转换，
->3264使用旧函数=使用旧函数）3265 self.\u num\u parallel\u calls=ops.convert\u to\u tensor（3266 num\u parallel\u calls，dtype=dtypes.int32，name=“num\u parallel\u calls”）
c:\users\toby pc\documents\code\blindness\u kaggle\my_env\lib\site packages\tensorflow\python\data\ops\dataset\ops.py in\uuuuu init\uuuuuuuuuuu（self，func，transformation\u name，dataset，input\u类，input\u形状，input\u类型，input\u结构，添加到图中，使用遗留函数，defun\ukwargs）2589 resource\u tracker=tracker.ResourceTracker（）2590带跟踪。资源跟踪者范围（资源跟踪者）：
->2591 self.\u function=wrapper\u fn.\u get\u concrete\u function\u internal（）2592如果将\u添加到图中：2593 self.\u function.将\u添加到图中（ops.get\u default\u graph（））
c:\users\toby pc\documents\code\blindness\u kaggle\my\u env\lib\site packages\tensorflow\python\eager\function.py in\u get\u concrete\u function\u internal（self，*args，**kwargs）1364”““获取图形函数时绕过错误检查。”“1365 graph\u function=self.\u get\u concrete\u function\u internal\u垃圾收集”(
->1366*args，**kwargs）1367#我们将这个具体函数返回给某人，他们可能保留对FuncGraph的1368#引用，而不保留对
c:\users\toby pc\documents\code\blindness\u kaggle\my\u env\lib\site packages\tensorflow\python\eager\function.py in\u get\u concrete\u function\u internal\u garbage\u collected（self，*args，
**kwargs）1358如果self.input_签名：1359 args，kwargs=None，None
->1360图形函数，可能定义函数（args，kwargs）1361返回图形函数1362
c:\users\toby pc\documents\code\blindness\u kaggle\my\u env\lib\site packages\tensorflow\python\eager\function.py in\u mayby\u define\u function（self，args，kwargs）1646 graph\u function=self.\u function\u cache.primary.get（cache\u key，None）1647如果graph\u function为None：
->1648 graph_function=self.\u create_graph_function（args，kwargs）1649 self.\u function_cache.primary[cache_key]=graph_function 1650返回graph_function，args，kwargs
c:\users\toby pc\documents\code\blindness\u kaggle\my_env\lib\site packages\tensorflow\python\eager\function.py in\u create\u graph\u函数（self、args、kwargs、override\u flat\u arg\u shapes）1539 arg\u name=arg\u name，1540 override\u flat\u arg\u arg\u shapes=override\u flat\u arg\u arg\u shapes，
->1541按值捕获=自身。按值捕获），1542自身。功能属性）1543
c:\users\toby pc\documents\code\blindness\u kaggle\my\u env\lib\site packages\tensorflow\python\framework\func\u graph.py来自func\u py func（名称、python\u func、args、kwargs、签名、函数图、自动签名、自动签名、自动签名选项、添加控制依赖项、参数名、op\u返回值、集合、按值捕获、覆盖平面arg\u形状）
714（转换函数）
715
-->716 func_outputs=python_func（*func_args，**func_kwargs）
717
718#不变量：`func_outputs`只包含张量、复合传感器、，
c:\users\toby pc\documents\code\blindness\u kaggle\my_env\lib\site packages\tensorflow\python\data\ops\dataset\ops.py在wrapper\u fn（*args）2583 attributes=defun\u kwargs）2584 def wrapper\u fn（*args）：\pylint:disable=缺少docstring
->2585 ret=_包装器_帮助器（*args）2586 ret=self._输出_结构._to_张量_列表（ret）2587返回[ops.convert_to_张量（t）for t in ret]
c:\users\toby pc\documents\code\blindness\u kaggle\my_env\lib\site packages\tensorflow\python\data\ops\dataset\ops.py in _wrapper\u helper（*args）2528嵌套的参数=（嵌套的参数，）2529
->2530 ret=func（*nested_args）2531#如果'func'返回一个张量列表，'nest.flatte（）`和2532#'ops.convert_to_tensor（）`将合谋尝试堆栈
在加载和预处理图像中（路径）
1 def加载和预处理图像（路径）：
2 image=tf.io.read\u文件（路径）
---->3返回预处理图像（图像）
在预处理_图像（图像）中
11水平比例系数=tf.数学减少平均值（行）
1.
import tensorflow as tf

def central_crop_tf(image, central_fraction):
    s = tf.shape(image)
    h, w = s[-3], s[-2]
    h_box = tf.cast(tf.round(central_fraction * tf.cast(h, tf.float32)), tf.int32)
    w_box = tf.cast(tf.round(central_fraction * tf.cast(w, tf.float32)), tf.int32)
    h_off = (h - h_box) // 2
    w_off = (w - w_box) // 2
    return tf.image.crop_to_bounding_box(image, h_off, w_off, h_box, w_box)

# Test
with tf.Graph().as_default(), tf.Session() as sess:
    img = tf.reshape(tf.range(80), [1, 8, 10, 1])
    print(sess.run(img)[0, :, :, 0])
    # [[ 0  1  2  3  4  5  6  7  8  9]
    #  [10 11 12 13 14 15 16 17 18 19]
    #  [20 21 22 23 24 25 26 27 28 29]
    #  [30 31 32 33 34 35 36 37 38 39]
    #  [40 41 42 43 44 45 46 47 48 49]
    #  [50 51 52 53 54 55 56 57 58 59]
    #  [60 61 62 63 64 65 66 67 68 69]
    #  [70 71 72 73 74 75 76 77 78 79]]
    frac = tf.constant(0.4)
    res = central_crop_tf(img, frac)
    print(sess.run(res)[0, :, :, 0])
    # [[23 24 25 26]
    #  [33 34 35 36]
    #  [43 44 45 46]]