zhongniuniu / PAMRFuse

Unsupervised fusion of misaligned PAT and MRI images via mutually reinforcing cross-modality image generation and registration
9 stars 1 forks source link

关于训练输入数据的格式问题 #3

Open yaoliu0803 opened 7 months ago

yaoliu0803 commented 7 months ago

作者你好,请教一下关于配准部分代码中的数据处理部分问题。 1.在配准训练的代码中,作者输入的是进行模拟形变的图像吗? image 这部分,我在配置文件中看超参数level设置的为0,是已经配准好的图像作为输入吗? 2.在上部分,进行的transform变换中,作者设置的了RandomAffine(),所以该网络能够进行配准的形变范围是怎样的?只能进行RandomAffine仿射配准吗? 3.我想使用自己的数据集去进行训练,在创建自己数据集的时候,需要进行哪些操作?(包括按照对应目录将为配准的图像放入相应文件夹,还需要将图片转为.npy吗?)除此之外,还需要进行哪些操作?

zhongniuniu commented 6 months ago
  1. since my input image is already misaligned, I don't need to simulate deformation and set the level to 0. This setting is for other people to use to simulate deformation in the aligned dataset.
  2. our alignment network performs elastic alignment.
  3. it is recommended that you first look at how the dataset is laid out and the format of the dataset provided by our code.