search

ZiJ-Wang / VQSRS

The official code for "Fast Real‑Time Brain Tumor Detection Based on Stimulated Raman Histology and Self‑Supervised Deep Learning Model".
3 stars 0 forks source link

Clarification on Preparing Training and Reconstructing Data for OpenSRH #1

Open MshGhazale opened 1 month ago

MshGhazale commented 1 month ago

Hello,

I recently downloaded the OpenSRH dataset and have some questions about preparing the training data:

Training Data: I understand that the training folder should contain color images. However, the patches in the OpenSRH dataset are grayscale, and there are two subfolders, CH2 and CH3, containing DICOM images. Should I subtract the images from these two subfolders to create a color image for training? If not, could you clarify the correct process for creating the training dataset from these grayscale images?

Reconstructing Data: Could you please explain what should be placed in the re_path folder? I'm unsure about its purpose and how to prepare data for it.

I would greatly appreciate your guidance as I’m currently a bit confused about the dataset preparation.

Thank you for your help!

ZiJ-Wang commented 1 month ago

Hi Msh,

You're correct that the OpenSRH dataset consists of grayscale images. As I mentioned in my paper, the dataset needs preprocessing before training. I’ve updated the preprocessing code in the data folder for your reference—please have a look.

Regarding the reconstruction data, you can place any preprocessed patch images in the re_path folder. In my work, I randomly selected one image from each of the seven tumor categories and placed them in the re_path folder. However, please ensure that the format follows the ImageNet dataset structure.

I hope this helps!

MshGhazale commented 1 month ago

Thank you very much for your prompt and detailed response. Your clarification was very helpful, and I appreciate your time and support.