学习日记【Eigen】：初遇Tensor_eigen tensor gpu

1.Eigen普通支持Matrix为二维矩阵，三维矩阵，或者更高维的，叫Tensor

2.对于深度学习训练还是想使用tensorflow

3.多维的Tensor矩阵操作起来还是很不方便，至少现在不太清楚是怎么操作的

4.在处理Tensor.mean时，数据过多且tensor为int类型时，可能存在总和越界，求出来为负值情况，而且只有tensor数据类型为double时，与opencv结果保持一致

学习过程中的部分测试代码:

#include "color.h"
void main()
{
       std::string image_path = "test_image/temp.bmp";
       Color* color = Color::Init();
       cv::Mat image;
       color->readImage(image_path, image);
       //image.convertTo(image, CV_32F,1.0/255.0);
       //cv::resize(image, image, cv::Size(0, 0), 0.5, 0.5);
       cv::cvtColor(image, image, cv::COLOR_BGR2RGB);
       Eigen::Tensor<double, 3> m_image(image.rows,image.cols,image.channels());
       cv::cv2eigen(image, m_image);
       // 对每个维度求均值
       //Eigen::Tensor<int, 0> mean_0 = m_image.mean(0);
       //Eigen::Tensor<int, 0> mean_1 = m_image.mean(1);
       //Eigen::Tensor<int, 0> mean_2 = m_image.mean(2);
       //std::cout << m_image << std::endl;
       std::cout << "mean 0: " << m_image.mean() << std::endl;
       //std::cout << "mean 1: " << mean_1(0) << std::endl;
       //std::cout << "mean 2: " << mean_2(0) << std::endl;
       std::cout << cv::mean(image) << std::endl;
       std::cout << (cv::mean(image)[0] + cv::mean(image)[1] +  cv::mean(image)[2])/3.0 << std::endl;
}