如何编写Keras代码，使用我训练好的模型读取我的自有图片？

Question

我使用 MNIST 训练了自己的数字识别模型，但当我尝试上传自己的图片进行预测时，它提示一个错误："ValueError: Input 0 of layer dense_3 is incompatible with the layer: expected axis -1 of input shape to have value 784 but received input with shape [None, 84]"（我的模型训练正确，用 MNIST 预测图片也成功）。以下是我的代码：

import numpy as np
from keras.applications.imagenet_utils import decode_predictions
from keras.preprocessing import image
from keras.applications import *
import glob
 
import os

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

img = []
x = []

images = image.load_img("/content/gdrive/My Drive/Colab Notebooks/num.png", target_size=(28, 28))
x = image.img_to_array(images)
x = np.expand_dims(x, axis=1)
img.append(x)

print(len(x))
x = np.concatenate([x for x in img])
 
model = tf.keras.models.load_model('num_reader.model')
y = model.predict(x)
print('Predicted:', decode_predictions(y, top=3))

这是我的错误:

28    #This is printed by "print(len(x))"
WARNING:tensorflow:Model was constructed with shape (None, 28, 28) for input Tensor("flatten_1_input_3:0", shape=(None, 28, 28), dtype=float32), but it was called on an input with incompatible shape (None, 1, 28, 3).
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-18-cd1af7600aac> in <module>()
     27 
     28 model = tf.keras.models.load_model('num_reader.model')
---> 29 y = model.predict(x)
     30 print('Predicted:', decode_predictions(y, top=3))

10 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/func_graph.py in wrapper(*args, **kwargs)
    971           except Exception as e:  # pylint:disable=broad-except
    972             if hasattr(e, "ag_error_metadata"):
--> 973               raise e.ag_error_metadata.to_exception(e)
    974             else:
    975               raise

ValueError: in user code:

    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:1462 predict_function  *
        return step_function(self, iterator)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:1452 step_function  **
        outputs = model.distribute_strategy.run(run_step, args=(data,))
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:1211 run
        return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:2585 call_for_each_replica
        return self._call_for_each_replica(fn, args, kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/distribute/distribute_lib.py:2945 _call_for_each_replica
        return fn(*args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:1445 run_step  **
        outputs = model.predict_step(data)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py:1418 predict_step
        return self(x, training=False)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer.py:985 __call__
        outputs = call_fn(inputs, *args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/sequential.py:372 call
        return super(Sequential, self).call(inputs, training=training, mask=mask)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/functional.py:386 call
        inputs, training=training, mask=mask)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/functional.py:508 _run_internal_graph
        outputs = node.layer(*args, **kwargs)
    /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/base_layer.py:976 __call__
        self.name)
    /usr/local/lib/python3.6/dist-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_3 is incompatible with the layer: expected axis -1 of input shape to have value 784 but received input with shape [None, 84]

Minst是一张28x28的图片，所以这段代码使用了28x28的图片。 我的开发环境是：Google Colab和Jupyter Notebook，并且我已经在两种环境下尝试过，仍然出现了这个错误。 有人可以帮忙吗？ 导入tensorflow库，它是一个深度学习库。张量只是多维数组。

mnist = tf.keras.datasets.mnist  # mnist is a dataset of 28x28 images of handwritten digits and their labels
(x_train, y_train),(x_test, y_test) = mnist.load_data()  # unpacks images to x_train/x_test and labels to y_train/y_test

x_train = tf.keras.utils.normalize(x_train, axis=1)  # scales data between 0 and 1
x_test = tf.keras.utils.normalize(x_test, axis=1)  # scales data between 0 and 1

model = tf.keras.models.Sequential()  # a basic feed-forward model
model.add(tf.keras.layers.Flatten())  # takes our 28x28 and makes it 1x784
model.add(tf.keras.layers.Dense(128, activation=tf.nn.relu))  # a simple fully-connected layer, 128 units, relu activation
model.add(tf.keras.layers.Dense(10, activation=tf.nn.softmax))  # our output layer. 10 units for 10 classes. Softmax for probability distribution

model.compile(optimizer='adam',  # Good default optimizer to start with
              loss='sparse_categorical_crossentropy',  # how will we calculate our "error." Neural network aims to minimize loss.
              metrics=['accuracy'])  # what to track

model.fit(x_train, y_train, epochs=10)  # train the model

val_loss, val_acc = model.evaluate(x_test, y_test)  # evaluate the out of sample data with model
print(val_loss)  # model's loss (error)
print(val_acc)  # model's accuracy

- 藍俊翔

Answer 1

不知道您的模型是如何构建的，以及num.png的具体情况，因此有些难以确定。但是，从错误消息来看，似乎您有两个错误：您正在加载rgb彩色图像(尽管MNIST和可能您的模型是在灰度图像上训练的)，因此(或者这是第二个错误)您的尺寸不正确(对于RGB是3，而不是灰度的1)。 28 * 3 = 84就是我解释错误消息的方式:

但接收到了形状为[None, 84]的输入。尝试添加rgb_to_grayscale和axis=0:

import tensorflow as tf
from keras_preprocessing import image
images = image.load_img("/content/gdrive/My Drive/Colab Notebooks/num.png", target_size=(28, 28))    
x = image.img_to_array(images)
x = tf.image.rgb_to_grayscale(x)
x = np.expand_dims(x, axis=0)
x = x/255.0
img.append(x)

print(len(x))
x = np.concatenate([x for x in img])
 
model = tf.keras.models.load_model('num_reader.model')
y = model.predict(x)
print('Predicted:', decode_predictions(y, top=3))

请注意，我添加了 x=x/255.0，因为我假设你忘记将值缩放到 [0,1] 的范围内。我假设如果你严格遵循 MNIST 机器学习教程，在预处理图像时已经对其进行了缩放。你也必须这样做。

更新: 使用将 x 直接传递给 model.predict，对我来说也可以正常工作：

images = image.load_img("/content/gdrive/My Drive/Colab Notebooks/num.png", target_size=(28, 28))    
x = image.img_to_array(images)
x = tf.image.rgb_to_grayscale(x)
x = np.expand_dims(x, axis=0)
x = x/255.0

model = tf.keras.models.load_model('num_reader.model')
model.predict(x)

- Stat Tistician

---

我已经添加了它，但仍然出现相同的错误。 - 藍俊翔

我刚才改了代码的顺序，抱歉，你能再检查一下吗？请确保tf.image.rgb_to_grayscale在图像加载之后。错误还是出现了吗？ - Stat Tistician

我已经添加了我想要预测的图片。 - 藍俊翔

好的，请检查更新后的代码，我测试通过了，没有错误。我修改了顺序和拼写错误。如果你发现错误，请告诉我。 - Stat Tistician

是的，它在图像加载后面，但仍然存在这个问题。 - 藍俊翔

Answer 2

检查输出层的神经元数量是否与类别数量匹配。

我曾经遇到过同样的问题，结果发现在二元分类中，模型的输出单元应该是一个而不是两个。

- glitch

Answer 3

在拟合模型之前，您需要使用to_categorical更改您的训练和测试标签。

from keras.utils import to_categorical
train_label=to_categorical(train_label)
test_labels=to_categorical(test_labels)

- Alperen Sayar