#include <caffe/caffe.hpp>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <iosfwd>
#include <memory>
#include <string>
#include <utility>
#include <vector>
#include <iostream>
//#define CPU_ONLY // 这里也可以在 项目属性/ C/C++ / 预定义选项中定义，没有装GPU电脑上，需要加上这个宏
using namespace caffe; // NOLINT(build/namespaces)
using std::string;
//using namespace System;
/* Pair (label, confidence) representing a prediction. */
typedef std::pair<string, float> Prediction;
class Classifier {
public:
Classifier(const string& model_file,
const string& trained_file,
const string& mean_file,
const string& label_file);
~Classifier();
std::vector<Prediction> Classify(const cv::Mat& img, int N = 5);
private:
void SetMean(const string& mean_file);// 从二进制的bin文件中读取均值，并设置到blob_中
std::vector<float> Predict(const cv::Mat& img); // 对图片进行预测
void WrapInputLayer(std::vector<cv::Mat>* input_channels);// 将net_的数据接口与input_channels 对接
void Preprocess(const cv::Mat& img,
std::vector<cv::Mat>* input_channels); // 以img为输入，用net_来forword计算输出层值。
private:
shared_ptr<Net<float> > net_; //网络对象
cv::Size input_geometry_; //输入数据的几何维度,宽和高
int num_channels_;//通道数
cv::Mat mean_;// 均值
std::vector<string> labels_; //各类的标记
};
Classifier::~Classifier() // 自己添加的函数，给的例程中是没有的
{
mean_.release();
labels_.clear();
}
Classifier::Classifier(const string& model_file,
const string& trained_file,
const string& mean_file,
const string& label_file) {
#ifdef CPU_ONLY
Caffe::set_mode(Caffe::CPU);
#else
Caffe::set_mode(Caffe::GPU);
#endif
std::cout << "set cpu" << std::endl;
/* Load the network. */
net_.reset(new Net<float>(model_file, TEST)); // 加载网络拓扑结构
net_->CopyTrainedLayersFrom(trained_file); // 加载网络权重
std::cout << "0" << std::endl;
CHECK_EQ(net_->num_inputs(), 1) << "Network should have exactly one input."; // 调用glog的检查
CHECK_EQ(net_->num_outputs(), 1) << "Network should have exactly one output.";// 检查
std::cout << "1" << std::endl;
Blob<float>* input_layer = net_->input_blobs()[0];
num_channels_ = input_layer->channels();
CHECK(num_channels_ == 3 || num_channels_ == 1)
<< "Input layer should have 1 or 3 channels.";
input_geometry_ = cv::Size(input_layer->width(), input_layer->height());//
std::cout << "2" << std::endl;
/* Load the binaryproto mean file. */
SetMean(mean_file);
std::cout << "3" << std::endl;
/* Load labels. */
std::ifstream labels(label_file.c_str());
CHECK(labels) << "Unable to open labels file " << label_file;
string line;
while (std::getline(labels, line))
labels_.push_back(string(line));
std::cout << "4" << std::endl;
Blob<float>* output_layer = net_->output_blobs()[0];
CHECK_EQ(labels_.size(), output_layer->channels())
<< "Number of labels is different from the output layer dimension.";//检查labels_的长度与输出层的维数是否一致
std::cout << "5" << std::endl;
}
// 下面两个函数是排序函数代码
static bool PairCompare(const std::pair<float, int>& lhs,
const std::pair<float, int>& rhs) {
return lhs.first > rhs.first;
}
/* Return the indices of the top N values of vector v. */
static std::vector<int> Argmax(const std::vector<float>& v, int N) {
std::vector<std::pair<float, int> > pairs;
for (size_t i = 0; i < v.size(); ++i)
pairs.push_back(std::make_pair(v[i], i));
std::partial_sort(pairs.begin(), pairs.begin() + N, pairs.end(), PairCompare);
std::vector<int> result;
for (int i = 0; i < N; ++i)
result.push_back(pairs[i].second);
return result;
}
/* Return the top N predictions. */
std::vector<Prediction> Classifier::Classify(const cv::Mat& img, int N) {
std::vector<float> output = Predict(img);
std::vector<int> maxN = Argmax(output, N);// 取前N个预测结果
std::vector<Prediction> predictions;
for (int i = 0; i < N; ++i) {
int idx = maxN[i];
predictions.push_back(std::make_pair(labels_[idx], output[idx]));//保存在predictions中
}
return predictions;
}
/* Load the mean file in binaryproto format. */
void Classifier::SetMean(const string& mean_file) {
BlobProto blob_proto; // 调用google/protobuf?? ,用于加速运算的数据接口，有时间再详细了解其应用方法
//这个函数是实现了从二进制文件中读取数据到blob_proto中,猜测函数来自第3方库的google/protobuf模块
ReadProtoFromBinaryFileOrDie(mean_file.c_str(), &blob_proto);
/* Convert from BlobProto to Blob<float> */
Blob<float> mean_blob;
mean_blob.FromProto(blob_proto); // 调用Blob类的成员函数FromRroto从BlobProto中加载数据
CHECK_EQ(mean_blob.channels(), num_channels_)
<< "Number of channels of mean file doesn't match input layer.";
/* The format of the mean file is planar 32-bit float BGR or grayscale. */
std::vector<cv::Mat> channels;
float* data = mean_blob.mutable_cpu_data();// 用可读写的方式取得指针
// 把均值上的各个通道的复制到 vector<Mat> channels，即channels[0]中对应均值中的通道0，
// 这样做的原因是 Blob类的数据存储方式是一维的。
// 我们这里是把一维度的数组 转化为Mat数组了
for (int i = 0; i < num_channels_; ++i) {
/* Extract an individual channel. */
cv::Mat channel(mean_blob.height(), mean_blob.width(), CV_32FC1, data);
channels.push_back(channel);
data += mean_blob.height() * mean_blob.width();
}
/* Merge the separate channels into a single image. */
cv::Mat mean;
cv::merge(channels, mean);//合并分开的通道为一个图像，即把channels的所有Mat合并为一个Mat.
/* Compute the global mean pixel value and create a mean image
* filled with this value. */
cv::Scalar channel_mean = cv::mean(mean); //计算每个像素在所有通道上的平均值，保存在channel_mean中
mean_ = cv::Mat(input_geometry_, mean.type(), channel_mean); //赋值给 本类的成员变量mean_
}
std::vector<float> Classifier::Predict(const cv::Mat& img) {
Blob<float>* input_layer = net_->input_blobs()[0]; // 得到net的输入层数据指针
input_layer->Reshape(1, num_channels_,
input_geometry_.height, input_geometry_.width);//分配内存？？？
/* Forward dimension change to all layers. */
net_->Reshape();
std::vector<cv::Mat> input_channels;
WrapInputLayer(&input_channels);// 将net_->input_blobs()[0]的地址给input_channels
Preprocess(img, &input_channels);//将图片地址给input_channels
net_->ForwardPrefilled(); //猜测是所有层前向计算
/* Copy the output layer to a std::vector */
// 将net的输出层数据复制到vector<float>类型的变量中，并返回
Blob<float>* output_layer = net_->output_blobs()[0];
const float* begin = output_layer->cpu_data();
const float* end = begin + output_layer->channels();
return std::vector<float>(begin, end);
}
/* Wrap the input layer of the network in separate cv::Mat objects
* (one per channel). This way we save one memcpy operation and we
* don't need to rely on cudaMemcpy2D. The last preprocessing
* operation will write the separate channels directly to the input
* layer. */
void Classifier::WrapInputLayer(std::vector<cv::Mat>* input_channels) {
Blob<float>* input_layer = net_->input_blobs()[0];
int width = input_layer->width();
int height = input_layer->height();
float* input_data = input_layer->mutable_cpu_data();// 取出Blob类的数据，并在后续部分对齐进行修改(即在Preprocess中，将图片的值放入input_layer中。
for (int i = 0; i < input_layer->channels(); ++i) {
cv::Mat channel(height, width, CV_32FC1, input_data);
input_channels->push_back(channel);
input_data += width * height;
}
}
void Classifier::Preprocess(const cv::Mat& img,
std::vector<cv::Mat>* input_channels) {
/* Convert the input image to the input image format of the network. */
// 保证输入图片的channels与 网络channels一致
cv::Mat sample;
if (img.channels() == 3 && num_channels_ == 1)
cv::cvtColor(img, sample, CV_BGR2GRAY);
else if (img.channels() == 4 && num_channels_ == 1)
cv::cvtColor(img, sample, CV_BGRA2GRAY);
else if (img.channels() == 4 && num_channels_ == 3)
cv::cvtColor(img, sample, CV_BGRA2BGR);
else if (img.channels() == 1 && num_channels_ == 3)
cv::cvtColor(img, sample, CV_GRAY2BGR);
else
sample = img;
// 保证大小一致
cv::Mat sample_resized;
if (sample.size() != input_geometry_)
cv::resize(sample, sample_resized, input_geometry_);
else
sample_resized = sample;
// 保证数据类型一致为 float
cv::Mat sample_float;
if (num_channels_ == 3)
sample_resized.convertTo(sample_float, CV_32FC3);
else
sample_resized.convertTo(sample_float, CV_32FC1);
// 减去均值得到sample_normalized
cv::Mat sample_normalized;
cv::subtract(sample_float, mean_, sample_normalized);
/* This operation will write the separate BGR planes directly to the
* input layer of the network because it is wrapped by the cv::Mat
* objects in input_channels. */
//将 sample_normalized 放入 input_channels中，即放入net_->input_blob中。
cv::split(sample_normalized, *input_channels);
CHECK(reinterpret_cast<float*>(input_channels->at(0).data)
== net_->input_blobs()[0]->cpu_data())
<< "Input channels are not wrapping the input layer of the network.";
}
int main(int argc, char** argv) {
try{
if (argc != 6) {
std::cerr << "Usage: " << argv[0]
<< " deploy.prototxt network.caffemodel"
<< " mean.binaryproto labels.txt img.jpg" << std::endl;
return 1;
}
::google::InitGoogleLogging(argv[0]);//glog 库内函数，glog 库是一个做日志的库
string model_file = argv[1];
string trained_file = argv[2];
string mean_file = argv[3];
string label_file = argv[4];
Classifier classifier(model_file, trained_file, mean_file, label_file);
string file = argv[5];
/* Load imglists. */
std::ifstream imglists(file.c_str());
CHECK(imglists) << "Unable to open labels file " << label_file;
string line;
while (std::getline(imglists, line))
{
//labels_.push_back(string(line));
string filename(line);
std::cout << "---------- Prediction for "
<< filename << " ----------" << std::endl;
cv::Mat img = cv::imread(filename, -1);
// 这里开始用图片列表，并且显示出来
CHECK(!img.empty()) << "Unable to decode image " << filename;
std::vector<Prediction> predictions = classifier.Classify(img, 2);
/* Print the top N predictions. */
//for (size_t i = 0; i < predictions.size(); ++i) {
for (size_t i = 0; i < 1; ++i) {
Prediction p = predictions[i];
// std::cout << std::fixed << std::setprecision(4) << p.second << " - \""
// << p.first << "\"" << std::endl;
std::cout << std::fixed << std::setprecision(4) << p.second << "," << p.first << std::endl;
}
CvFont font;
cvInitFont(&font, CV_FONT_HERSHEY_DUPLEX, 1.0f, 1.0f, 0, 1, CV_AA);
//cv::addText(img, predictions[1].first, cv::Point(10, 10), &font);
for (int i = 6; i < 7; i++)
{
cv::Mat imgt = img.clone();
cv::putText(imgt, predictions[0].first, cv::Point(80, 40), i, 2.0f, CV_RGB(255, 0, 0));
cv::imshow("img", imgt);
cv::waitKey(1);
}
}
std::cout << "done" << std::endl;
classifier.~Classifier();
return 0;
exit(0);
}
//catch (ArithmeticException^ e)
//{
// Console::WriteLine("ArithmeticException Handler: {0}", e);
//}
//catch (Exception^ e)
//{
// Console::WriteLine("Generic Exception Handler: {0}", e);
//}
catch (std::exception e)
{
std::cout << e.what() << std::endl;
}
}