一、背景
原本是打算按《DEX Deep EXpectation of apparent age from a single image》进行表面年龄的训练,可由于IMDB-WIKI的数据集比较庞大,各个年龄段分布不均匀,难以划分训练集及验证集。后来为了先跑通整个训练过程的主要部分,就直接用LAP数据集,参考caffe的finetune_flickr_style,进行一些参数修改,利用bvlc_reference_caffenet.caffemodel完成年龄估计的finetune。
二、训练数据集准备
1、下载LAP数据集,包括Train、Validation、Test,以及对应的年龄label,http://chalearnlap.cvc.uab.es/dataset/18/description/,需要注册。也可以从我的网盘下载:
链接:https://pan.baidu.com/s/1536TgbR_cCcS7-QHfEAeMw
提取码:xc45
2、将标注好的csv文件转换为caffe识别的txt格式。csv每一行的信息为:图片名、年龄、标准差。训练的时候不需要标准差信息,我们只要将图片名和年龄写入到txt中,并按空格隔开,得到Train.txt如下:
同样,完成验证集cvs文件的转换,得到Validation.txt。
三、模型及相关文件拷贝
1、拷贝预训练好的vgg16模型caffe\models\bvlc_reference_caffenet\bvlc_reference_caffenet.caffemodel至工作目录下,该文件约232M;
2、拷贝caffe\models\finetune_flickr_style文件夹中deploy.prototxt、solver.prototxt、train_val.prototxt至工作目录下;
3、拷贝imageNet的均值文件caffe\data\ilsvrc12\imagenet_mean.binaryproto至工作目录下。
四、参数修改
1、修改train_val.prototxt
以及最后的输出层个数,因为我们要训练的为[0,100]岁的输出,共101类,所以:
2、修改solver.protxt
3、修改用于实际测试的部署文件deploy.protxt
输出层的个数也要改:
五、开始训练
1、新建train.bat
caffe train -solver solver.prototxt -weights bvlc_reference_caffenet.caffemodel rem caffe train --solver solver.prototxt --snapshot snapshot/bvlc_iter_48000.solverstate pause
双击即可开始训练,当训练过程中出现意外中断,可注释第一行,关闭第二行注释,根据实际情况修改保存,继续双击训练。
我的电脑CPU是i5 6500,显卡为gtx1050Ti,8G内存,大致要训练10个小时吧,中途也出现了一些内存不足训练终止的情况。
2、训练结束
六、模型评价
年龄估计原本是一个线性问题,不是一个明确的分类问题,人都无法准确无误地得到某人的年龄,更何况是机器呢。所以评价这个年龄分类模型的好坏不能简单地通过精度来衡量,可以用MAE(平均绝对误差)以及ε-error来衡量,其中
1、对验证集Validation.txt的所有图片进行预测
借助 https://github.com/eveningglow/age-and-gender-classification ,其环境搭建可参考https://www.cnblogs.com/smbx-ztbz/p/9399016.html
修改main函数
int split(std::string str, std::string pattern, std::vector<std::string> &words) { std::string::size_type pos; std::string word; int num = 0; str += pattern; std::string::size_type size = str.size(); for (auto i = 0; i < size; i++) { pos = str.find(pattern, i); if (pos == i) { continue;//if first string is pattern } if (pos < size) { word = str.substr(i, pos - i); words.push_back(word); i = pos + pattern.size() - 1; num++; } } return num; } //param example: model/deploy_age2.prototxt model/age_net.caffemodel model/mean.binaryproto img/0008.jpg int main(int argc, char* argv[]) { if (argc != 5) { cout << "Command shoud be like ..." << endl; cout << "AgeAndGenderClassification "; cout << " \"AGE_NET_MODEL_FILE_PATH\" \"AGE_NET_WEIGHT_FILE_PATH\" \"MEAN_FILE_PATH\" \"TEST_IMAGE\" " << endl; std::cout << "argc = " << argc << std::endl; getchar(); return 0; } // Get each file path string age_model(argv[1]); string age_weight(argv[2]); string mean_file(argv[3]); //string test_image(argv[4]); // Probability vector vector<Dtype> prob_age_vec; // Set mode Caffe::set_mode(Caffe::GPU); // Make AgeNet AgeNet age_net(age_model, age_weight, mean_file); // Initiailize both nets age_net.initNetwork(); //读取待测试的图片名 std::ifstream fin("E:\\caffe\\DEX_age_gender_predict\\lap2\\Validation.txt"); std::string line; std::vector<std::string> test_images; std::vector<int> test_images_age; while (!fin.eof()) { std::getline(fin, line); std::vector<std::string> words; split(line, " ", words); test_images.push_back(words[0]); test_images_age.push_back(atoi(words[1].c_str())); } std::cout << "test_images size = " << test_images.size() << std::endl; std::ofstream fout("E:\\caffe\\DEX_age_gender_predict\\lap2\\Validation_predict.txt"); for (int k = 0; k < test_images.size(); ++k) { std::cout << "k = " << k << std::endl; std::string test_image; test_image = test_images[k]; // Classify and get probabilities Mat test_img = imread(test_image, CV_LOAD_IMAGE_COLOR); int age = age_net.classify(test_img, prob_age_vec); // Print result and show image //std::cout << "prob_age_vec size = " << prob_age_vec.size() << std::endl; //for (int i = 0; i < prob_age_vec.size(); ++i) { // std::cout << "[" << i << "] = " << prob_age_vec[i] << std::endl; //} //Dtype prob; //int index; //get_max_value(prob_age_vec, prob, index); //std::cout << "prob = " << prob << ", index = " << index << std::endl; //imshow("AgeAndGender", test_img); //waitKey(0); fout << test_images[k] << " " << test_images_age[k] << " " << age << std::endl; } std::cout << "finish!" << std::endl; getchar(); return 0; }
我的命令参数为:E:\caffe\DEX_age_gender_predict\lap2\deploy.prototxt E:\caffe\DEX_age_gender_predict\lap2\snapshot\bvlc_iter_50000.caffemodel model\mean.binaryproto img\0008.jpg
可根据实际情况修改。可得到Validation_predict.txt文件。运行过程中可能会因为内存不足中断运行,可能要分批次运行多次。
2、计算MAE及ε-error
(1)将Validation_predict.txt文件及验证集的标注文件Reference.csv拷贝到新建的vs项目的工作目录下;
(2)计算
#include <iostream> #include <string> #include <fstream> #include <vector> int split(std::string str, std::string pattern, std::vector<std::string> &words) { std::string::size_type pos; std::string word; int num = 0; str += pattern; std::string::size_type size = str.size(); for (auto i = 0; i < size; i++) { pos = str.find(pattern, i); if (pos == i) { continue;//if first string is pattern } if (pos < size) { word = str.substr(i, pos - i); words.push_back(word); i = pos + pattern.size() - 1; num++; } } return num; } int main(int argc, char** argv) { //u, sigma, x std::vector<int> u; std::vector<float> sigma; std::vector<int> predict; std::string line; std::ifstream csv_file("Reference.csv"); while (!csv_file.eof()) { std::getline(csv_file, line); std::vector<std::string> words; split(line, ";", words); u.push_back(atoi(words[1].c_str())); sigma.push_back(atof(words[2].c_str())); } std::ifstream predict_file("Validation_predict.txt"); while (!predict_file.eof()) { std::getline(predict_file, line); std::vector<std::string> words; split(line, " ", words); predict.push_back(atoi(words[2].c_str())); } if (u.size() != predict.size()) { std::cout << "u.size() != predict.size()" << std::endl; getchar(); return -1; } //MAE int sum_err = 0; float MAE = 0; for (int i = 0; i < u.size(); ++i) { sum_err += abs(u[i] - predict[i]); } MAE = static_cast<float>(sum_err) / u.size(); std::cout << "MAE = " << MAE << std::endl;//11.7184 //esro-error std::vector<float> errors; float err = 0; float error = 0.0; for (int i = 0; i < u.size(); ++i) { err = 1.0 - exp(-1.0*(predict[i] - u[i])*(predict[i] - u[i]) / (2 * sigma[i] * sigma[i])); errors.push_back(err); error += err; } error /= errors.size(); std::cout << "error = " << error << std::endl;//0.682652 std::cout << "finish!" << std::endl; getchar(); return 0; }
最终得到MAE为11.7184, ε-error为0.682652。
七、实际应用中预测
1、可利用caffe提供的classification工具对输入图片地进行估计
classification deploy.prototxt snapshot\bvlc_iter_50000.caffemodel imagenet_mean.binaryproto ..\age_labels.txt ..\test_image\test_3.jpg
pause
其中,age_labels.txt为0-100个label的说明信息,每个label对应一行,共101行,我的写法如下:
end