The requirements for running project are listed in requirements.txt. To install the requirements, run:pip install -r requirements.txt.
Download the weights and test data and extract them to the project directory.
weight&&test_data,passwd:w133
verse2020 datasets verse2020.
data of Guizhou Orthopaedic Hospital (not yet public).
In order to unify the data, every image needs to be reoriented to RAI.The transformation script is in other/preprocess.py, and set the input folder, output folder and then run it.
Also need to transform coordinate information by other/process_landmark_2020.py.
Set task, save_dir, dataset_dir in the data_generator.py script to generate training data for the corresponding task.
The training scripts are in the train folder, e.g. To train spine segmentation,run:python train/train_spine_segment.py --save_dir spine_save --train_dataset_dir spine_train --test_dataset_dir spine_test --batch_size 8 --n_epoch 100.
See the script file for more details.
Change the reference_image_path and save_path in the test.py script to reason about a single CT image.
Set postprocessing to enable postprocessing.
Set single_channel to select a different localization model.
See the script file for more details.
Change the gt_base_folder, pred_base_folder, and output_folder in the eval.py script to calculate the localization and segmentation metrics.
See the script file for more details.
visualization tools 3D Slicer.
