LBP人脸识别的python实现

这几天看了看LBP及其人脸识别的流程，并在网络上搜相应的python代码，有，但代码质量不好，于是自己就重新写了下，对于att_faces数据集的识别率能达到95.0%~99.0%（40种类型，每种随机选5张训练，5张识别），全部代码如下，不到80行哦。

#coding:utf-8

import numpy as np

import cv2, os, math, os.path, glob, random

g_mapping=[

    0, 1, 2, 3, 4, 58, 5, 6, 7, 58, 58, 58, 8, 58, 9, 10,

    11, 58, 58, 58, 58, 58, 58, 58, 12, 58, 58, 58, 13, 58, 14, 15,

    16, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    17, 58, 58, 58, 58, 58, 58, 58, 18, 58, 58, 58, 19, 58, 20, 21,

    22, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    23, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    24, 58, 58, 58, 58, 58, 58, 58, 25, 58, 58, 58, 26, 58, 27, 28,

    29, 30, 58, 31, 58, 58, 58, 32, 58, 58, 58, 58, 58, 58, 58, 33,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 34,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 35,

    36, 37, 58, 38, 58, 58, 58, 39, 58, 58, 58, 58, 58, 58, 58, 40,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 41,

    42, 43, 58, 44, 58, 58, 58, 45, 58, 58, 58, 58, 58, 58, 58, 46,

    47, 48, 58, 49, 58, 58, 58, 50, 51, 52, 58, 53, 54, 55, 56, 57]

def loadImageSet(folder, sampleCount=5):

    trainData = []; testData = []; yTrain=[]; yTest = [];

    for k in range(1,41):

        folder2 = os.path.join(folder, 's%d' %k)

        data = [cv2.imread(d.encode('gbk'),0) for d in glob.glob(os.path.join(folder2, '*.pgm'))]

        sample = random.sample(range(10), sampleCount)

        trainData.extend([data[i] for i in range(10) if i in sample])

        testData.extend([data[i] for i in range(10) if i not in sample])

        yTest.extend([k]* (10-sampleCount))

        yTrain.extend([k]* sampleCount)

    return trainData, testData, np.array(yTrain), np.array(yTest)

def LBP(I, radius=2, count=8):       #得到图像的LBP特征

    dh = np.round([radius*math.sin(i*2*math.pi/count) for i in range(count)])

    dw = np.round([radius*math.cos(i*2*math.pi/count) for i in range(count)])

    height ,width = I.shape

    lbp = np.zeros(I.shape, dtype = np.int)

    I1 = np.pad(I, radius, 'edge')

    for k in range(count):

        h,w = radius+dh[k], radius+dw[k]

        lbp += ((I>I1[h:h+height, w:w+width])<<k)

    return lbp

def calLbpHistogram(lbp, hCount=7, wCount=5, maxLbpValue=255): #分块计算lbp直方图

    height,width = lbp.shape

    res = np.zeros((hCount*wCount, max(g_mapping)+1), dtype=np.float)

    assert(maxLbpValue+1 == len(g_mapping))

    for h  in range(hCount):

        for w in range(wCount):

            blk = lbp[height*h/hCount:height*(h+1)/hCount, width*w/wCount:width*(w+1)/wCount]

            hist1 = np.bincount(blk.ravel(), minlength=maxLbpValue)

            hist = res[h*wCount+w,:]

            for v,k in zip(hist1, g_mapping):

                hist[k] += v

            hist /= hist.sum()

    return res

def main(folder=u'E:/迅雷下载/faceProcess/att_faces'):

    trainImg, testImg, yTrain, yTest = loadImageSet(folder)

    xTrain = np.array([calLbpHistogram(LBP(d)).ravel() for d in trainImg])

    xTest  = np.array([calLbpHistogram(LBP(d)).ravel() for d in testImg])

    lsvc = cv2.SVM()                              #支持向量机方法

    svm_params = dict( kernel_type = cv2.SVM_LINEAR, svm_type = cv2.SVM_C_SVC, C=2.67, gamma=5.383 )

    lsvc.train(np.float32(xTrain), np.float32(yTrain), params = svm_params)

    lsvc_y_predict = np.array( [lsvc.predict(d) for d in np.float32(xTest)])

    print u'支持向量机识别率', (lsvc_y_predict == np.array(yTest)).mean()  

if __name__ == '__main__':

    main()

　　下面是对mnist手写数字数据集的识别，修改了数据集的载入，并加了图像的倾斜校正，识别率达到96%（如果使用sklearn的svm，效率会更高一些。）

import cPickle

import gzip,math

import numpy as np

import os, glob, random, cv2

SZ = 28

def deskew(img):

    m = cv2.moments(img)

    if abs(m['mu02']) < 1e-2:

        return img.copy()

    skew = m['mu11']/m['mu02']

    M = np.float32([[1, skew, -0.5*SZ*skew], [0, 1, 0]])

    img = cv2.warpAffine(img,M,(SZ, SZ),flags=cv2.WARP_INVERSE_MAP|cv2.INTER_LINEAR)

    return img

g_mapping=[

    0, 1, 2, 3, 4, 58, 5, 6, 7, 58, 58, 58, 8, 58, 9, 10,

    11, 58, 58, 58, 58, 58, 58, 58, 12, 58, 58, 58, 13, 58, 14, 15,

    16, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    17, 58, 58, 58, 58, 58, 58, 58, 18, 58, 58, 58, 19, 58, 20, 21,

    22, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    23, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    24, 58, 58, 58, 58, 58, 58, 58, 25, 58, 58, 58, 26, 58, 27, 28,

    29, 30, 58, 31, 58, 58, 58, 32, 58, 58, 58, 58, 58, 58, 58, 33,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 34,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 35,

    36, 37, 58, 38, 58, 58, 58, 39, 58, 58, 58, 58, 58, 58, 58, 40,

    58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 58, 41,

    42, 43, 58, 44, 58, 58, 58, 45, 58, 58, 58, 58, 58, 58, 58, 46,

    47, 48, 58, 49, 58, 58, 58, 50, 51, 52, 58, 53, 54, 55, 56, 57]

def loadImageSet():

    with gzip.open('./mnist.pkl.gz') as fp:

        train_set, valid_set, test_set = cPickle.load(fp)

    xTrain = train_set[0]; s1 = xTrain.shape; xTrain = xTrain.reshape((s1[0],28,28))

    xTest = test_set[0];   s2 = xTest.shape;  xTest = xTest.reshape((s2[0],28,28))

    xTrain = np.array([deskew(d) for d in xTrain])

    xTest  = np.array([deskew(d) for d in xTest])

    return xTrain, xTest, train_set[1],  test_set[1]

def LBP(I, radius=2, count=8):       #得到图像的LBP特征

    dh = np.round([radius*math.sin(i*2*math.pi/count) for i in range(count)])

    dw = np.round([radius*math.cos(i*2*math.pi/count) for i in range(count)])

    height ,width = I.shape

    lbp = np.zeros(I.shape, dtype = np.int)

    I1 = np.pad(I, radius, 'edge')

    for k in range(count):

        h,w = radius+dh[k], radius+dw[k]

        lbp += ((I>I1[h:h+height, w:w+width])<<k)

    return lbp

def calLbpHistogram(lbp, hCount=2, wCount=2, maxLbpValue=255): #分块计算lbp直方图

    height,width = lbp.shape

    res = np.zeros((hCount*wCount, max(g_mapping)+1), dtype=np.float)

    assert(maxLbpValue+1 == len(g_mapping))

    for h  in range(hCount):

        for w in range(wCount):

            blk = lbp[height*h/hCount:height*(h+1)/hCount, width*w/wCount:width*(w+1)/wCount]

            hist1 = np.bincount(blk.ravel(), minlength=maxLbpValue)

            hist = res[h*wCount+w,:]

            for v,k in zip(hist1, g_mapping):

                hist[k] += v

            hist /= hist.sum()

    return res

def main():

    trainImg, testImg, yTrain, yTest = loadImageSet()

    xTrain = np.array([calLbpHistogram(LBP(d)).ravel() for d in trainImg])

    xTest  = np.array([calLbpHistogram(LBP(d)).ravel() for d in testImg])

    lsvc = cv2.SVM()                              #支持向量机方法

    svm_params = dict( kernel_type = cv2.SVM_LINEAR, svm_type = cv2.SVM_C_SVC, C=2.67, gamma=5.383 )

    lsvc.train(np.float32(xTrain), np.float32(yTrain), params = svm_params)

    lsvc_y_predict = np.array( [lsvc.predict(d) for d in np.float32(xTest)])

    print u'支持向量机', (lsvc_y_predict == np.array(yTest)).mean()  

if __name__ == '__main__':

    main()

LBP人脸识别的python实现的更多相关文章

gabor变换人脸识别的python实现，att_faces数据集平均识别率99%
大家都说gabor做人脸识别是传统方法中效果最好的,这几天就折腾实现了下,网上的python实现实在太少,github上的某个版本还误导了我好几天,后来采用将C++代码封装成dll供python调用的 ...
PCA人脸识别的python实现
这几天看了看PCA及其人脸识别的流程,并在网络上搜相应的python代码,有,但代码质量不好,于是自己就重新写了下,对于att_faces数据集的识别率能达到92.5%~98.0%(40种类型,每种随 ...
iOS活体人脸识别的Demo和一些思路
代码地址如下:http://www.demodashi.com/demo/12011.html 之前公司项目需要,研究了一下人脸识别和活体识别,并运用免费的讯飞人脸识别,在其基础上做了二次开发,添加了 ...
人脸检测? 对Python来说太简单, 调用dlib包就可以完成
"Dlib 是一个现代化的 C ++ 工具包,包含用于创建复杂软件的机器学习算法和工具 " .它使您能够直接在 Python 中运行许多任务,其中一个例子就是人脸检测. 安装 dl ...
百度Aip人脸识别之python代码
用python来做人脸识别代码量少思路清晰, 在使用之前我们需要在我们的配置的编译器中通过pip install baidu-aip 即可 from aip import AipFace 就可以开 ...
人脸识别之Python DLib库进行人脸关键点识别
一.首先安装DLib模块这里只介绍linux安装的过程,windows安装过程请自行百度 1.首先,安装dlib.skimage前:先安装libboost sudo apt-get install ...
转《在浏览器中使用tensorflow.js进行人脸识别的JavaScript API》
作者 | Vincent Mühle 编译 | 姗姗出品 | 人工智能头条(公众号ID:AI_Thinker) [导读]随着深度学习方法的应用,浏览器调用人脸识别技术已经得到了更广泛的应用与提升.在 ...
face_recognition 人脸识别报错
[root@localhost examples]# python facerec_from_video_file.py RuntimeError: module compiled against A ...
face-api.js：一个在浏览器中进行人脸识别的 JavaScript 接口
Mark! 本文将为大家介绍一个建立在「tensorflow.js」内核上的 javascript API——「face-api.js」,它实现了三种卷积神经网络架构,用于完成人脸检测.识别和特征点检 ...

随机推荐

python已写内容中可能的报错及解决办法
理论上我发的每个短文,直接复制放到py里面,python xx.py是可以执行的,不过因为版本,编码什么的问题会有报错,详见这里报错: SyntaxError: Non-ASCII characte ...
固定UILabel宽度分行显示
固定UILabel宽度分行显示这种小伎俩估计都被用烂了,笔者给大家提供一个category文件,供大家简单设置哦. 各种富文本效果哦(普通文本也是可以用的呢): 3行,固定宽度200 2行,固定宽度 ...
Python初学者第二十二天函数进阶（1）
22day 1.函数命名空间: 2.函数作用域的查找顺序:LEGB locals->enclosing function ->globals ->_builtins_ a.local ...
wrap装饰器包
import wrapt # without argument in decorator @wrapt.decorator def logging(wrapped, instance, args, k ...
Hibernate核心对象
1.Configuration Configuration 类负责管理Hibernate的配置信息.它包括如下内容: Hibernate运行的底层信息:数据库的URL.用户名.密码.JDBC驱动类,数 ...
SQL触发器与CLR的使用
在数据库的日常操作中,面对复杂业务的情况下,总会有用sql语句或存储过程不是那么方便的时候,所以这时候就会想到在数据库中调用CLR,也就是调用程序集,此处用C#实现来讲解一个测试案例测试案例的业务是 ...
[装]JMX监控Hadoop
http://chenjc-it.iteye.com/blog/1539746 实验成功!
ant jmeter jenkins 实现自动化测试
编者按: 本文核心的关键是ant去调用build.xml 执行自动化测试任务. bulid.xml 文件的内容按照注释修改为你自己服务器相关信息. 配置完成后一定要手动在服务器上执行一次,确保你配置的 ...
Hive学习之路（二十）Hive 执行过程实例分析
一.Hive 执行过程概述 1.概述 (1) Hive 将 HQL 转换成一组操作符(Operator),比如 GroupByOperator, JoinOperator 等 (2)操作符 Opera ...
Hadoop学习之路（三）Hadoop-2.7.5在CentOS-6.7上的编译
下载Hadoop源码 1.登录官网 2.确定你要安装的软件的版本一个选取原则: 不新不旧的稳定版本几个标准: 1)一般来说,刚刚发布的大版本都是有很多问题 2)应该选择某个大版本中的最后一个小版本 ...

LBP人脸识别的python实现

LBP人脸识别的python实现的更多相关文章

随机推荐

热门专题