用于多类分类交叉熵损失函数的 Keras CNN

Question

这是我输入的代码，用于 class 验证一些 class 由鸟、狗和猫组成的内容。它与二进制 classification 的代码相同但是当我添加另一个 class 并将编译方法的损失函数更改为使用 categorical_Crossentropy 时，它给了我以下错误（ => 在代码结束）。谁能解释这里的问题或我犯的错误？

# Importing Keras and Tensorflow modules

import tensorflow as tf
from keras.models import Sequential
from keras.layers import Conv2D
from keras.layers import MaxPooling2D
from keras.layers import Flatten
from keras.layers import Dense
from keras.preprocessing.image import ImageDataGenerator
from keras.models import load_model
from keras.utils.np_utils import to_categorical
import os.path

# Initilize the CNN

classifier = Sequential()
# Step 1 - Convolution

classifier.add(Conv2D(32, (3, 3), input_shape = (64, 64, 3), activation = 'relu'))

# Step 2 - Pooling

classifier.add(MaxPooling2D(pool_size = (2, 2)))

# Step 2(b) - Add 2nd Convolution Layer making it Deep followed by a Pooling Layer

classifier.add(Conv2D(32, (3, 3), activation = 'relu'))
classifier.add(MaxPooling2D(pool_size = (2, 2)))

# Step 3 - Flattening

classifier.add(Flatten())

# Step 4 - Fully Connected Neural Network

# Hidden Layer - Activation Function RELU
classifier.add(Dense(units = 128, activation = 'relu')) 
# Output Layer - Activation Function Softmax(to clasify multiple classes)
classifier.add(Dense(units = 1, activation = 'softmax'))

# Compile the CNN

# Categorical Crossentropy - to classify between multiple classes of images
classifier.compile(optimizer = 'adam', loss = 'categorical_crossentropy', 
metrics = ['accuracy'])

# Image Augmentation and Training Section

# Image Augmentation to prevent Overfitting (Applying random transformation on 
images to train set.ie. 
# scalling, rotating and streching)

train_datagen = ImageDataGenerator(
        rescale=1./255,
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True)

test_datagen = ImageDataGenerator(rescale=1./255)

training_set = train_datagen.flow_from_directory(
        'dataset/training_set',
        target_size=(64, 64),
        batch_size=8,
        class_mode='categorical')

test_set = test_datagen.flow_from_directory(
        'dataset/test_set',
        target_size=(64, 64),
        batch_size=8,
        class_mode='categorical')

#Fit the clasifier on the CNN data
if(os.path.isfile('my_model.h5') == False):
    classifier.fit_generator(
            training_set,
            steps_per_epoch=8000,
            epochs=2,
            validation_data=test_set,
            validation_steps=2000
                                  )
# Save the generated model to my_model.h5
classifier.save('my_model.h5')
else:
    classifier = load_model('my_model.h5')

Answer 1

在最后一个（密集）层上，每个 class 需要一个神经元。

classifier.add(Dense(3))

现在你只有一个神经元，而且你的网络仍然只设置两个 classes。

Answer 2

您的数据集似乎有 3 个类，因此您需要将模型定义中的最后一行更改为：

classifier.add(Dense(units = 3, activation = 'softmax'))

用于多类分类交叉熵损失函数的 Keras CNN

Keras CNN for multiclass categorical crossentropy loss function

python

convolution

deep-learning

keras

tensorflow