itk::Image和libtorch::tensor 互相转换的代码实现

最新推荐文章于 2023-08-15 10:36:09 发布

lizy_fish

最新推荐文章于 2023-08-15 10:36:09 发布

阅读量1.2k

点赞数 2

分类专栏： ITK记录文章标签： c++ pytorch

本文链接：https://blog.csdn.net/lizy_fish/article/details/107812822

版权

方法一：itk::image 转buffer,然后调用torch::from_blob

步骤如下：
1，读取图像到itk::Image
2，将itk::Image 的数据写入连续的buffer中
3，调用libtorch的from_blob函数将buffer的内容转入tensor

// read image with itk
    const char* im_path = "H:\\tmp\\a\\src.mhd";
    using ImageType = itk::Image<float,3>;
    using ReaderType = itk::ImageFileReader<ImageType>;
    ReaderType::Pointer reader = ReaderType::New();
    reader->SetFileName(im_path);
    ImageType::Pointer image = reader->GetOutput();
    reader->Update();

    // print some point for debug
    std::cout<<"For itk Image3d:"<<std::endl;
    std::cout<<"point[224,297,10] must be 2086, now is "
            <<image->GetPixel({
   224,297,10})<<std::endl;
    std::cout<<"point[113,285,10] must be 167.3, now is "
            <<image->GetPixel({
   113,285,10})<<std::endl;
    std::cout<<"point[301,371,5] must be 1086, now is "
            <<image->GetPixel({
   301,371,5})<<std::endl