Keras Multiclass Classification input_shape 大小问题
Keras Multiclass Clasification input_shape size problem
我正在使用 Keras 构建人脸检测 AI。我收集了以下形式的数据：
数据集
- 一种
-b
- C
-d
main.py
其中，a、b、c、d 是 类 个，每个都包含图像。我想使用 image_dataset_from_directory 命令，但是当我尝试将所有图像的大小调整为 400x400 时，它会提示我形状 160000 和 400x400 不兼容。但是，如果我将图像大小调整为 1,160000，则一切正常，但由于图像失真，准确性很差。感谢任何帮助。
我正在使用 python 3.7 与 tensorflow 2.3 和 tf-nightly
这是错误：
WARNING:tensorflow:Model was constructed with shape (None, 400, 400) for input Tensor("flatten_input:0", shape=(None, 400, 400), dtype=float32), but it was called on an input with incompatible shape (400, 400, 1).
Traceback (most recent call last):
File "C:\Users\PycharmProjects\custom face\mainm.py", line 48, in <module>
prediction = model.predict_on_batch(pred[1][0])
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1766, in predict_on_batch
outputs = predict_function(iterator)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 780, in __call__
result = self._call(*args, **kwds)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 823, in _call
self._initialize(args, kwds, add_initializers_to=initializers)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 697, in _initialize
*args, **kwds))
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\function.py", line 2841, in _get_concrete_function_internal_garbage_collected
graph_function, _, _ = self._maybe_define_function(args, kwargs)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\function.py", line 3199, in _maybe_define_function
graph_function = self._create_graph_function(args, kwargs)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\function.py", line 3061, in _create_graph_function
capture_by_value=self._capture_by_value),
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\framework\func_graph.py", line 979, in func_graph_from_py_func
func_outputs = python_func(*func_args, **func_kwargs)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 600, in wrapped_fn
return weak_wrapped_fn().__wrapped__(*args, **kwds)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\framework\func_graph.py", line 966, in wrapper
raise e.ag_error_metadata.to_exception(e)
ValueError: in user code:
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1440 predict_function *
return step_function(self, iterator)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1430 step_function **
outputs = model.distribute_strategy.run(run_step, args=(data,))
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:1063 run
return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2377 call_for_each_replica
return self._call_for_each_replica(fn, args, kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2736 _call_for_each_replica
return fn(*args, **kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1423 run_step **
outputs = model.predict_step(data)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1396 predict_step
return self(x, training=False)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\base_layer.py:961 __call__
outputs = call_fn(inputs, *args, **kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\sequential.py:372 call
return super(Sequential, self).call(inputs, training=training, mask=mask)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\functional.py:385 call
inputs, training=training, mask=mask)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\functional.py:507 _run_internal_graph
outputs = node.layer(*args, **kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\base_layer.py:952 __call__
self.name)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\input_spec.py:216 assert_input_compatibility
' but received input with shape ' + str(shape))
ValueError: Input 0 of layer dense is incompatible with the layer: expected axis -1 of input shape to have value 160000 but received input with shape [400, 400]
这是我的代码：
from libraries import *
tf.compat.v1.enable_eager_execution()
gpu_options = tf.compat.v1.GPUOptions(per_process_gpu_memory_fraction=0.9)
sess = tf.compat.v1.Session(config=tf.compat.v1.ConfigProto(gpu_options=gpu_options))
data = tf.keras.preprocessing.image_dataset_from_directory("dataset", labels="inferred", label_mode="categorical",class_names=['a','b','c','d'], color_mode='grayscale', image_size=(400,400), )
labels = ['k','am','ash','an']
model = keras.Sequential([ #sequential says to list layers in sequence
keras.layers.Flatten(input_shape=(400,400)), #Input layer setup. making data from [[1],[2],[3]] to [1,2,3]
keras.layers.Dense(128, activation="relu"), #1st hidden layer of 128 neurons, Dense means all layers interconnected equally. Activation is the rectifier linear
# unit making the proccesing complex. It does this by making all negetives 0 and all positives bigger.
keras.layers.Dense(128, activation="relu"),
keras.layers.Dense(100, activation="relu"),
keras.layers.Dense(4, activation="softmax")#softmaxout layer of 10 neurons, activation gives a probability for each of the 10 output choices
])
# Compile up
model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])# loss function is where you look at the probability and
# change weights and biases based on how much you lost. metric is what you want to achieve.
#Train
checkpoint_path = "training_1/cp.ckpt"
checkpoint_dir = os.path.dirname(checkpoint_path)
# Create a callback that saves the model's weights
cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_path,
save_weights_only=True,
verbose=1)
#model.fit(data,epochs=1, callbacks = [cp_callback]) #epochs are how many times the network sees the same data during training. feeds data randomly. gives
model.load_weights("training_1/cp.ckpt")
test_loss, test_acc = model.evaluate(data)
print("Tested Loss: ", test_loss)
print("Tested Acc: ", test_acc*100, "%")
unbatched = data.unbatch()
print("================")
pred = list(unbatched.as_numpy_iterator())
print(len(pred))
prediction = model.predict_on_batch(pred[1][0])
# img = Image.fromarray(pred[0][0], 'RGB')
# img.save('my.png')
# img.show()
print(prediction)
keras.layers.Flatten(...
您不能使用该层启动模型，它只接受向量。
使用 Conv2D，然后是下一层。如果您能看到更多工作模型的示例，那是最好的。这将使您获得更好的结果并减少错误。
我正在使用 Keras 构建人脸检测 AI。我收集了以下形式的数据：
数据集 - 一种 -b - C -d main.py
其中，a、b、c、d 是 类 个，每个都包含图像。我想使用 image_dataset_from_directory 命令，但是当我尝试将所有图像的大小调整为 400x400 时，它会提示我形状 160000 和 400x400 不兼容。但是，如果我将图像大小调整为 1,160000，则一切正常，但由于图像失真，准确性很差。感谢任何帮助。
我正在使用 python 3.7 与 tensorflow 2.3 和 tf-nightly
这是错误：
WARNING:tensorflow:Model was constructed with shape (None, 400, 400) for input Tensor("flatten_input:0", shape=(None, 400, 400), dtype=float32), but it was called on an input with incompatible shape (400, 400, 1).
Traceback (most recent call last):
File "C:\Users\PycharmProjects\custom face\mainm.py", line 48, in <module>
prediction = model.predict_on_batch(pred[1][0])
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1766, in predict_on_batch
outputs = predict_function(iterator)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 780, in __call__
result = self._call(*args, **kwds)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 823, in _call
self._initialize(args, kwds, add_initializers_to=initializers)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 697, in _initialize
*args, **kwds))
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\function.py", line 2841, in _get_concrete_function_internal_garbage_collected
graph_function, _, _ = self._maybe_define_function(args, kwargs)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\function.py", line 3199, in _maybe_define_function
graph_function = self._create_graph_function(args, kwargs)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\function.py", line 3061, in _create_graph_function
capture_by_value=self._capture_by_value),
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\framework\func_graph.py", line 979, in func_graph_from_py_func
func_outputs = python_func(*func_args, **func_kwargs)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\eager\def_function.py", line 600, in wrapped_fn
return weak_wrapped_fn().__wrapped__(*args, **kwds)
File "C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\framework\func_graph.py", line 966, in wrapper
raise e.ag_error_metadata.to_exception(e)
ValueError: in user code:
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1440 predict_function *
return step_function(self, iterator)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1430 step_function **
outputs = model.distribute_strategy.run(run_step, args=(data,))
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:1063 run
return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2377 call_for_each_replica
return self._call_for_each_replica(fn, args, kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\distribute\distribute_lib.py:2736 _call_for_each_replica
return fn(*args, **kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1423 run_step **
outputs = model.predict_step(data)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\training.py:1396 predict_step
return self(x, training=False)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\base_layer.py:961 __call__
outputs = call_fn(inputs, *args, **kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\sequential.py:372 call
return super(Sequential, self).call(inputs, training=training, mask=mask)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\functional.py:385 call
inputs, training=training, mask=mask)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\functional.py:507 _run_internal_graph
outputs = node.layer(*args, **kwargs)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\base_layer.py:952 __call__
self.name)
C:\Users\PycharmProjects\custom face\venv\lib\site-packages\tensorflow\python\keras\engine\input_spec.py:216 assert_input_compatibility
' but received input with shape ' + str(shape))
ValueError: Input 0 of layer dense is incompatible with the layer: expected axis -1 of input shape to have value 160000 but received input with shape [400, 400]
这是我的代码：
from libraries import *
tf.compat.v1.enable_eager_execution()
gpu_options = tf.compat.v1.GPUOptions(per_process_gpu_memory_fraction=0.9)
sess = tf.compat.v1.Session(config=tf.compat.v1.ConfigProto(gpu_options=gpu_options))
data = tf.keras.preprocessing.image_dataset_from_directory("dataset", labels="inferred", label_mode="categorical",class_names=['a','b','c','d'], color_mode='grayscale', image_size=(400,400), )
labels = ['k','am','ash','an']
model = keras.Sequential([ #sequential says to list layers in sequence
keras.layers.Flatten(input_shape=(400,400)), #Input layer setup. making data from [[1],[2],[3]] to [1,2,3]
keras.layers.Dense(128, activation="relu"), #1st hidden layer of 128 neurons, Dense means all layers interconnected equally. Activation is the rectifier linear
# unit making the proccesing complex. It does this by making all negetives 0 and all positives bigger.
keras.layers.Dense(128, activation="relu"),
keras.layers.Dense(100, activation="relu"),
keras.layers.Dense(4, activation="softmax")#softmaxout layer of 10 neurons, activation gives a probability for each of the 10 output choices
])
# Compile up
model.compile(optimizer="adam", loss="categorical_crossentropy", metrics=["accuracy"])# loss function is where you look at the probability and
# change weights and biases based on how much you lost. metric is what you want to achieve.
#Train
checkpoint_path = "training_1/cp.ckpt"
checkpoint_dir = os.path.dirname(checkpoint_path)
# Create a callback that saves the model's weights
cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_path,
save_weights_only=True,
verbose=1)
#model.fit(data,epochs=1, callbacks = [cp_callback]) #epochs are how many times the network sees the same data during training. feeds data randomly. gives
model.load_weights("training_1/cp.ckpt")
test_loss, test_acc = model.evaluate(data)
print("Tested Loss: ", test_loss)
print("Tested Acc: ", test_acc*100, "%")
unbatched = data.unbatch()
print("================")
pred = list(unbatched.as_numpy_iterator())
print(len(pred))
prediction = model.predict_on_batch(pred[1][0])
# img = Image.fromarray(pred[0][0], 'RGB')
# img.save('my.png')
# img.show()
print(prediction)
keras.layers.Flatten(...
您不能使用该层启动模型，它只接受向量。 使用 Conv2D，然后是下一层。如果您能看到更多工作模型的示例，那是最好的。这将使您获得更好的结果并减少错误。