search

ZerojumpLine / CoLab

[TMI2023] Context Label Learning: Improving Background Class Representations in Semantic Segmentation.
MIT License
19 stars 2 forks source link

Data preprocessing #3

Open flying-lby opened 11 months ago

flying-lby commented 11 months ago

Hello, your work is fantastic, but I have a few questions. I hope you can provide me with some advice.

  1. How can I transform the dataset saved in the datapreprocessed folder into the dataset inside the datafiles folder?
  2. The README file only provides preprocessing code for the Liver Tumor data. Could you please advise on how to preprocess the other datasets used in the project?

Looking forward to your response. Thank you!

ZerojumpLine commented 10 months ago

Hi,

Sorry for the late response. I am trying my best to help you reimplement our method.

  1. We generate the datapath of the processed data and save to the datafiles, which make it easier in network training/inference. Now I add the code to generate those files at the end of datapreprocessing.ipynb, e.g. "Step 5, Generate Datafiles". Hope this can help you generate the datapath for your own datasets.
  2. We provide the download link for other datasets. You might refer to our other projects for the preprocessing code of the other datasets. For example, kidney tumor segmentation, and MRI preprocessing (which should be similar to use for brain tumor/lesion segmentation). Please note you need to generate the binary mask yourself as the preprocessing scripts I mention are developed for other applications.

Best, Zeju