keras 数据增强 cnn 总是期望分类器输出中有一个神经元？

Question

我正在实现一个具有数据增强功能的两类 cnn，但是除非输出层由单个神经元组成，否则网络总是会一直引发“预期形状错误”。这是我的代码。

from keras.models import Sequential
from keras.models import Model
from keras.layers import Dense, Activation
from keras.preprocessing.image import ImageDataGenerator
from keras.optimizers import Adam
from keras.layers import Dense, Conv2D, MaxPooling2D , Flatten, Dropout
import keras
from keras import losses
from keras import backend as K
from keras.utils import to_categorical
from matplotlib.pyplot import imread, imshow, subplots, show
import numpy as np


def my_model(input_shape, opt, no_of_class):
    model = Sequential()
    model.add(Conv2D(32, (3, 3), activation='relu', input_shape=input_shape))
    model.add(Conv2D(32, (3, 3), activation='relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Conv2D(64, (3, 3), activation='relu'))
    model.add(Conv2D(64, (3, 3), activation='relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Flatten())
    model.add(Dense(256, activation='relu'))
    model.add(Dense(no_of_class, activation='softmax'))
    model.compile(optimizer=opt, loss='binary_crossentropy', metrics=['accuracy'])
    model.summary()
    return model

PATH="C:\\Users\\****\\Desktop\\****\\****\\data\\poroxity\\"
train_data=PATH+'train'
validation_data=PATH+'validation'
no_of_class=2
training_samples=251
validation_samples=108
epochs=50
batch_size=16
img_width, img_height = 224, 224
opt = Adam(lr=0.001)

if K.image_data_format() == 'channels_first':
    input_shape = (3, img_width, img_height)
else:
    input_shape = (img_width, img_height, 3)

train_datagen=ImageDataGenerator(
        rotation_range=90,
        vertical_flip=True,
        width_shift_range=0.3, fill_mode='wrap',
        rescale=1./255,
        shear_range=0.2,
        horizontal_flip=True)
validation_datagen=ImageDataGenerator(rescale=1./255)

# validation and training iterators preparing loaded images from the directory
train_it = train_datagen.flow_from_directory(
    train_data,
    target_size=(img_height, img_width),
    class_mode='binary',
    batch_size=batch_size,
    shuffle=True,
    color_mode='rgb',
    seed=42,
    save_to_dir=PATH+'augmented',
    save_prefix='poroxity_aug',
    save_format='jpeg'
    )
validation_it = validation_datagen.flow_from_directory(
    validation_data,
    target_size=(img_height, img_width),
    class_mode='binary',
    batch_size=1)

model=my_model(input_shape, opt, no_of_class)
model.fit_generator(
        generator=train_it,
        epochs=epochs,
        verbose=1,
        steps_per_epoch=train_it.n // train_it.batch_size,
        validation_data=validation_it,
        validation_steps=validation_it.n // validation_it.batch_size,
        )

ValueError: 检查目标时出错：预期dense_2 的形状为(2,)，但得到的数组的形状为(1,)

一些输出

找到属于 2 个类别的 251 张图片。 找到属于 2 个类别的 108 个图像。 训练批次形状：(16, 224, 224, 3) 标签批次形状：(16,) 批次中的标签：[0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 1. 1. 0. 0. 1.] 验证批次形状：(1, 224, 224, 3) 标签批次形状：(1,) 批次中的标签：[0.] 输入形状：(224, 224, 3)

Answer 1

我通过将 model.compile 中的 loss 设置为 'categorical_crossentropy' 和 class_mode 在 train 的 .flow_from_directory 和验证中设置为 'categorical' 解决了这个问题。