Keras实现CIFAR-10分类

仅仅为了学习Keras的使用，使用一个四层的全连接网络对MNIST数据集进行分类，网络模型各层结点数为：3072: : 1024 : 512:10；

使用50000张图片进行训练，10000张测试：

              precision    recall  f1-score   support

    airplane       0.61      0.69      0.65      1000

  automobile       0.69      0.67      0.68      1000

        bird       0.43      0.49      0.45      1000

         cat       0.40      0.32      0.36      1000

        dear       0.49      0.50      0.50      1000

         dog       0.45      0.48      0.47      1000

        frog       0.58      0.65      0.61      1000

       horse       0.63      0.60      0.62      1000

        ship       0.72      0.66      0.69      1000

       truck       0.63      0.58      0.60      1000

   micro avg       0.56      0.56      0.56     10000

   macro avg       0.56      0.56      0.56     10000

weighted avg       0.56      0.56      0.56     10000

训练过程中，损失和正确率曲线：

可以看到，训练集的损失在一直降低，而测试集的损失出现大范围波动，并趋于上升，说明在一些epoch之后，出现过拟合；

训练集的正确率也在一直上升，并接近100%；而测试集的正确率达到50%就趋于平稳了；

代码：

#!/usr/bin/env python

# -*- coding: utf-8 -*-

# @Time : 19-5-9

"""

implement classification for CIFAR-10 with Keras

"""

__author__ = 'Zhen Chen'

# import the necessary packages

from sklearn.preprocessing import LabelBinarizer

from sklearn.metrics import classification_report

from keras.models import Sequential

from keras.layers import Dense

from keras.optimizers import SGD

from keras.datasets import cifar10

import matplotlib.pyplot as plt

import numpy as np

import argparse

# construct the argument parse and parse the arguments

parser = argparse.ArgumentParser()

parser.add_argument("-o", "--output", default="./Training Loss and Accuracy_CIFAR10.png")

args = parser.parse_args()

# load the training and testing data, scale it into the range [0, 1],

# then reshape the design matrix

print("[INFO] loading CIFAR-10 data...")

((trainX, trainY), (testX, testY)) = cifar10.load_data()

trainX = trainX.astype("float") / 255.0

testX = testX.astype("float") / 255.0

trainX = trainX.reshape((trainX.shape[0], 3072))

testX = testX.reshape((testX.shape[0], 3072))

# convert the labels from integers to vectors

lb = LabelBinarizer()

trainY = lb.fit_transform(trainY)

testY = lb.fit_transform(testY)

# initialize the label names for the CIFAR-10 dataset

labelNames = ["airplane", "automobile", "bird", "cat", "dear", "dog", "frog", "horse", "ship", "truck"]

# define the 2072-1024-512-10 architecture Keras

model = Sequential()

model.add(Dense(1024, input_shape=(3072,), activation="relu"))

model.add(Dense(512, activation="relu"))

model.add(Dense(10, activation="softmax"))

# train the model using SGD

print("[INFO] training network...")

sgd = SGD(0.01)

model.compile(loss="categorical_crossentropy", optimizer=sgd, metrics=["accuracy"])

H = model.fit(trainX, trainY, validation_data=(testX, testY), epochs=100, batch_size=32)

model.save_weights('SGD_100_32.h5')

# evaluate the network

print("[INFO] evaluating network...")

predictions = model.predict(testX, batch_size=32)

print(classification_report(testY.argmax(axis=1), predictions.argmax(axis=1), target_names=labelNames))

# plot the training losss and accuracy

plt.style.use("ggplot")

plt.figure()

plt.plot(np.arange(0, 100), H.history["loss"], label="train_loss")

plt.plot(np.arange(0, 100), H.history["val_loss"], label="val_loss")

plt.plot(np.arange(0, 100), H.history["acc"], label="train_acc")

plt.plot(np.arange(0, 100), H.history["val_acc"], label="val_acc")

plt.title("Training Loss and Accuracy on CIRFAR-10")

plt.xlabel("Epoch #")

plt.ylabel("Loss/Accuracy")

plt.legend()

plt.savefig(args.output)