Docker Tutorial of SegRap2025

This repository provides a template to build Docker for SegRap2025 final submission, which you can use as a starting point to develop your own final submitted algorithm.

Usage

• Put the weight of used model and other things in the resources folder, or anywhere else, as long as they are used correctly in the code. If you don't put the used files into resources folder, you should change the codes in the corresponding Dockerfile correctly to ensure the used files are correctly copied.
• Change the inference code (func generate_segmentation or something else) in infer.py.
• Write the required Python libraries in the requirements.txt file with corret version.

[NOTE] We recommend that the participants should run test_docker.sh of the two tasks before submission to ensure the correct building of your docker.

Description of the File Structure During the Final Test

The test set of Task01: GTV segmentation consist of paired CT and ceCT images from two cohorts, and the image are organized as follows:

test
 ├──input
 |   ├──Cohort1
 |   |   └──image
 |   |       ├──segrap_****
 |   |       |   ├──image.nii.gz
 |   |       |   └──image_contrast.nii.gz
 |   |       └──...
 |   └──Cohort2
 |       └──image
 |           ├──segrap_****
 |           |   ├──image.nii.gz
 |           |   └──image_contrast.nii.gz
 |           └──...
 └──output

The test set of Task02: LN CTV Segmentation consist of three kinds of cases: paired CT and ceCT (segrap_****), only CT (segrap_nc_****), and only ceCT (segrap_ce_****).

test
 ├──input
 |   └──image
 |       ├──segrap_****
 |       |   ├──image.nii.gz
 |       |   └──image_contrast.nii.gz
 |       ├──segrap_nc_****
 |       |   └──image.nii.gz
 |       ├──segrap_ce_****
 |       |   └──image_contrast.nii.gz
 |       └──...
 └──output

Challenge and Dataset

For any information about the challenge, please visit the official website.