机器学习笔记 - 使用dlib进行训练对猫脸进行识别

数据准备

是从猫脸关键点检测大赛下载的猫的数据集，训练集和测试集各上万张图片，不光含有图片，还有标记好的猫脸关键点的数据（用于关键点检测），我们先用图片自行打标记自行识别。

数据集百度网盘下载地址

链接：https://pan.baidu.com/s/1CjaAXx1yqse6Xf6sNG4Tkg 
提取码：098k 

图片标记

使用imglab进行标记，可参考https://blog.csdn.net/bashendixie5/article/details/110338937

训练集我标记了216张，测试集标记了41张。

训练代码

调整了多次C参数，不知道是因为图片数量的问题还是什么问题，C参数从5、10、15、20、25、30，准确率递增，由于没有使用gpu版本的dlib，有些慢，所以没有再增加。30的时候识别率大概再70%。猜测如果调整图片数量、质量、训练/测试数量比，应该能提高上来。

# 猫脸检测

import os
import sys
import glob
import dlib
import cv2

# options用于设置训练的参数和模式
options = dlib.simple_object_detector_training_options()
# Since faces are left/right symmetric we can tell the trainer to train a
# symmetric detector.  This helps it get the most value out of the training
# data.
options.add_left_right_image_flips = True
# 支持向量机的C参数，通常默认取为5.自己适当更改参数以达到最好的效果
options.C = 30
# 线程数，你电脑有4核的话就填4
options.num_threads = 8
options.be_verbose = True

# 获取路径
# train_folder = 'E:/dlib-master/tools/imglab/build/Release/'
# test_folder = current_path + '/cats_test/'
train_xml_path = 'E:/dlib-master/tools/imglab/build/Release/mydataset.xml'
test_xml_path = 'E:/dlib-master/tools/imglab/build/Release/testdataset.xml'

# print("training file path:" + train_xml_path)
# print(train_xml_path)
# print("testing file path:" + test_xml_path)
# print(test_xml_path)

# 开始训练
print("start training:")
dlib.train_simple_object_detector(train_xml_path, 'detector.svm', options)

#print("")  # Print blank line to create gap from previous output
print("Training accuracy: {}".format(
    dlib.test_simple_object_detector(train_xml_path, "detector.svm")))

print("Testing accuracy: {}".format(
    dlib.test_simple_object_detector(test_xml_path, "detector.svm")))

测试代码

import os
import sys
import dlib
import cv2
import glob

detector = dlib.simple_object_detector("detector.svm")

#current_path = os.getcwd()
test_folder = 'E:/dlib-master/tools/imglab/build/Release/addimages/'

for f in glob.glob(test_folder+'*.jpg'):
    print("Processing file: {}".format(f))
    img = cv2.imread(f, cv2.IMREAD_COLOR)
    b, g, r = cv2.split(img)
    img2 = cv2.merge([r, g, b])
    dets = detector(img2)
    print("Number of faces detected: {}".format(len(dets)))
    for index, face in enumerate(dets):
        print('face {}; left {}; top {}; right {}; bottom {}'.format(index, face.left(), face.top(), face.right(), face.bottom()))

        left = face.left()
        top = face.top()
        right = face.right()
        bottom = face.bottom()
        cv2.rectangle(img, (left, top), (right, bottom), (0, 255, 0), 3)
        cv2.namedWindow(f, cv2.WINDOW_AUTOSIZE)
        cv2.imshow(f, img)

k = cv2.waitKey(0)
cv2.destroyAllWindows()

测试结果，测试了19张图片，找到了8张，还可以。