model.predict抛出ValueError:预期的shape=(无，64，64，3)，发现shape=(无，64，3)

Question

我在google笔记本中使用tensorflow建立了以下卷积神经网络(CNN)模型：

import tensorflow as tf

PATCHSIZE = 64

# CNN model
model = tf.keras.models.Sequential([
    # Convolutional layer. Learn 32 filters using a 3x3 kernel
    tf.keras.layers.Conv2D(32, (3, 3), activation="relu", input_shape=(PATCHSIZE, PATCHSIZE, 3)),

    #Max-pooling layer, using 2x2 pool size
    tf.keras.layers.MaxPooling2D(pool_size=(2, 2)),

    # Flatten units
    tf.keras.layers.Flatten(),
    # Add a hidden layer with dropout
    tf.keras.layers.Dense(128, activation="relu"),
    tf.keras.layers.Dropout(0.5),

    # Add an output layer with one output unit for a probability
    tf.keras.layers.Dense(1, activation="sigmoid")
])

model.compile(
    optimizer="adam",
    loss="binary_crossentropy",
    metrics=["accuracy"]
)

training_patches = []
training_labels = []

training_patches = np.array(training_patches)
training_labels = np.array(training_labels)

model.fit(training_patches, training_labels, epochs=20)

validation_patches = np.array(validation_patches)
validation_labels = np.array(validation_labels)

model.evaluate(validation_patches, validation_labels, verbose=2)

输入数据(补丁)是大小为64x64像素的彩色图像。因此输入形状是(64，64，3)。训练和推理过程运行良好，但当我试图预测输入修补程序(形状为(64、64、3)的numpy数组)的概率时，我得到了以下错误：

throws  ValueError: Input 0 of layer "sequential_6" is incompatible with the layer: expected shape=(None, 64, 64, 3), found shape=(None, 64, 3)

 # throws  ValueError: Input 0 of layer "sequential_6" is incompatible with the layer: expected shape=(None, 64, 64, 3), found shape=(None, 64, 3)
model.predict(patch)

我做错了什么？

Answer 1

问题是您没有指定批处理维度。试试下面的代码：

import tensorflow as tf
import numpy as np

PATCHSIZE = 64

# CNN model
model = tf.keras.models.Sequential([
    # Convolutional layer. Learn 32 filters using a 3x3 kernel
    tf.keras.layers.Conv2D(32, (3, 3), activation="relu", input_shape=(PATCHSIZE, PATCHSIZE, 3)),

    #Max-pooling layer, using 2x2 pool size
    tf.keras.layers.MaxPooling2D(pool_size=(2, 2)),

    # Flatten units
    tf.keras.layers.Flatten(),
    # Add a hidden layer with dropout
    tf.keras.layers.Dense(128, activation="relu"),
    tf.keras.layers.Dropout(0.5),

    # Add an output layer with one output unit for a probability
    tf.keras.layers.Dense(1, activation="sigmoid")
])

model.compile(
    optimizer="adam",
    loss="binary_crossentropy",
    metrics=["accuracy"]
)

training_patches = np.ones((1, PATCHSIZE, PATCHSIZE, 3))
training_labels = np.ones((1, 1))

model.fit(training_patches, training_labels, epochs=20)

validation_patches = np.ones((1, PATCHSIZE, PATCHSIZE, 3))
validation_labels = np.ones((1, 1))

model.evaluate(validation_patches, validation_labels, verbose=2)

如果您有一个validation_patch，那么向已经存在的numpy数组添加维度的简单方法如下：

my_array = np.ones((64, 64, 3))
print(my_array.shape)  # output: (64, 64, 3)
my_array = np.expand_dims(my_array, axis=0)
print(my_array.shape)  # output: (1, 64, 64, 3)