用深度学习keras的cnn做图像识别分类，准确率达97%

 向AI转型的程序员都关注了这个号???

大数据挖掘DT数据分析  公众号： datadw

Keras是一个简约，高度模块化的神经网络库。

可以很容易和快速实现原型（通过总模块化，极简主义，和可扩展性）
同时支持卷积网络（vision）和复发性的网络（序列数据）。以及两者的组合。
无缝地运行在CPU和GPU上。

keras的资源库网址为https://github.com/fchollet/keras

olivettifaces人脸数据库介绍

Olivetti Faces是纽约大学的一个比较小的人脸库，由 40个人的400张图片构成，即每个人的人脸图片为10张。每张图片的灰度级为8位，每个像素的灰度大小位于0-255之间，每张图片大小为64×64。 如下图，这个图片大小是1140942，一共有2020张人脸，故每张人脸大小是（1140/20）（942/20）即5747=2679：

预处理模块

使用了PIL（Python Imaging Library）模块，是Python平台事实上的图像处理标准库。
预处理流程是：打开文件-》归一化-》将图片转为数据集-》生成label-》使用pickle序列化数据集

numpy.ndarray.flatten函数的功能是将一个矩阵平铺为向量

from PIL import Image

import numpy

import cPickle

img = Image.open('G:\data\olivettifaces.gif')

# numpy supports conversion from image to ndarray and normalization by dividing 255

# 1140 * 942 ndarray

img_ndarray = numpy.asarray(img, dtype='float64') / 255

# create numpy array of 400*2679

img_rows, img_cols = 57, 47

face_data = numpy.empty((400, img_rows*img_cols))

# convert 1140*942 ndarray to 400*2679 matrix

for row in range(20):

    for col in range(20):

        face_data[row*20+col] = numpy.ndarray.flatten(img_ndarray[row*img_rows:(row+1)*img_rows, col*img_cols:(col+1)*img_cols])

# create label

face_label = numpy.empty(400, dtype=int)

for i in range(400):

    face_label[i] = i / 10

# pickling file

f = open('G:\data\olivettifaces.pkl','wb')

# store data and label as a tuple

cPickle.dump((face_data,face_label), f)

f.close()

分类模型

程序参考了官方示例：https://github.com/fchollet/keras/blob/master/examples/mnist_cnn.py
一共有40个类，每个类10个样本，共400个样本。其中320个样本用于训练，40个用于验证，剩下40个测试
注意给第一层指定input_shape，如果是MLP，代码为：

model = Sequential()

# Dense(64) is a fully-connected layer with 64 hidden units.

# in the first layer, you must specify the expected input data shape:

# here, 20-dimensional vectors.

model.add(Dense(64, input_dim=20, init='uniform'))

后面可以不指定Dense的input shape

from __future__ import print_function

import numpy as np

import cPickle

np.random.seed(1337) # for reproducibililty

from keras.datasets import mnist

from keras.models import Sequential

from keras.layers.core import Dense, Dropout, Activation, Flatten

from keras.layers.convolutional import Convolution2D, MaxPooling2D

from keras.utils import np_utils

# split data into train,vavlid and test

# train:320

# valid:40

# test:40

def split_data(fname):

    f = open(fname, 'rb')

    face_data,face_label = cPickle.load(f)

    X_train = np.empty((320, img_rows * img_cols))

    Y_train = np.empty(320, dtype=int)

    X_valid = np.empty((40, img_rows* img_cols))

    Y_valid = np.empty(40, dtype=int)

    X_test = np.empty((40, img_rows* img_cols))

    Y_test = np.empty(40, dtype=int)

    for i in range(40):

        X_train[i*8:(i+1)*8,:] = face_data[i*10:i*10+8,:]

        Y_train[i*8:(i+1)*8] = face_label[i*10:i*10+8]

        X_valid[i] = face_data[i*10+8,:]

        Y_valid[i] = face_label[i*10+8]

        X_test[i] = face_data[i*10+9,:]

        Y_test[i] = face_label[i*10+9]

    

    return (X_train, Y_train, X_valid, Y_valid, X_test, Y_test)

if __name__=='__main__':

    batch_size = 10

    nb_classes = 40

    nb_epoch = 12

    # input image dimensions

    img_rows, img_cols = 57, 47

    # number of convolutional filters to use

    nb_filters = 32

    # size of pooling area for max pooling

    nb_pool = 2

    # convolution kernel size

    nb_conv = 3

    (X_train, Y_train, X_valid, Y_valid, X_test, Y_test) = split_data('G:\data\olivettifaces.pkl')

    X_train = X_train.reshape(X_train.shape[0], 1, img_rows, img_cols)

    X_test = X_test.reshape(X_test.shape[0], 1, img_rows, img_cols)

    

    print('X_train shape:', X_train.shape)

    print(X_train.shape[0], 'train samples')

    print(X_test.shape[0], 'test samples')

    # convert label to binary class matrix

    Y_train = np_utils.to_categorical(Y_train, nb_classes)

    Y_test = np_utils.to_categorical(Y_test, nb_classes)

    model = Sequential()

    # 32 convolution filters , the size of convolution kernel is 3 * 3

    # border_mode can be 'valid' or 'full'

    #‘valid’only apply filter to complete patches of the image. 

    # 'full'  zero-pads image to multiple of filter shape to generate output of shape: image_shape + filter_shape - 1

    # when used as the first layer, you should specify the shape of inputs 

    # the first number means the channel of an input image, 1 stands for grayscale imgs, 3 for RGB imgs

    model.add(Convolution2D(nb_filters, nb_conv, nb_conv,

                            border_mode='valid',

                            input_shape=(1, img_rows, img_cols)))

    # use rectifier linear units : max(0.0, x)

    model.add(Activation('relu'))

    # second convolution layer with 32 filters of size 3*3

    model.add(Convolution2D(nb_filters, nb_conv, nb_conv))

    model.add(Activation('relu'))

    # max pooling layer, pool size is 2 * 2

    model.add(MaxPooling2D(pool_size=(nb_pool, nb_pool)))

    # drop out of max-pooling layer , drop out rate is 0.25 

    model.add(Dropout(0.25))

    # flatten inputs from 2d to 1d

    model.add(Flatten())

    # add fully connected layer with 128 hidden units

    model.add(Dense(128))

    model.add(Activation('relu'))

    model.add(Dropout(0.5))

    # output layer with softmax 

    model.add(Dense(nb_classes))

    model.add(Activation('softmax'))

    # use cross-entropy cost and adadelta to optimize params

    model.compile(loss='categorical_crossentropy', optimizer='adadelta')

    # train model with bath_size =10, epoch=12

    # set verbose=1 to show train info

    model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch,

          show_accuracy=True, verbose=1, validation_data=(X_test, Y_test))

    # evaluate on test set

    score = model.evaluate(X_test, Y_test, show_accuracy=True, verbose=0)

    print('Test score:', score[0])

    print('Test accuracy:', score[1])

结果：
准确率有97%

via : http://www.cnblogs.com/wacc/p/5341654.html

人工智能大数据与深度学习

搜索添加微信公众号：weic2c

长按图片，识别二维码，点关注

---

大数据挖掘DT数据分析

搜索添加微信公众号：datadw

教你机器学习，教你数据挖掘

长按图片，识别二维码，点关注