Furen Xiao
a516408c4e
- Implemented 09_run_iterative.py for MONAI to orchestrate iterative pseudo-labeling rounds, including training, predicting, and evaluation. - Created 01_prepare_dataset.py, 02_plan_preprocess.py, 06_pseudo_label.py, 07_train.py, 08_eval.py, and 09_run_iterative.py for nnU-Net to manage dataset preparation, planning, training, evaluation, and iterative pseudo-labeling. - Enhanced error handling and logging throughout the scripts for better traceability. - Added functionality to handle negative pseudo-labels and TTA options during inference. - Included visualization of Dice scores across rounds for both MONAI and nnU-Net.
31 lines
No EOL
1.2 KiB
Python
31 lines
No EOL
1.2 KiB
Python
"""Build/rebuild the nnU-Net raw dataset (Dataset210_NTUH_T1C_PL) from rows jsonl.
|
|
|
|
rows: {key, img, label} where img is the NATIVE (unprocessed) T1c nifti and
|
|
label a nifti mask on (or warpable to) that grid. No resampling/cropping/
|
|
normalization is applied here — nnUNetv2_plan_and_preprocess (02) does the
|
|
preprocessing; imagesTr entries are symlinks to the native volumes, labelsTr
|
|
entries are symlinks when already 0/1 on the image grid, else nearest-warped
|
|
+ binarized copies (label fix-up only).
|
|
|
|
Usage: python scripts_nnu/01_nnu_prepare_dataset.py --rows <rows.jsonl>
|
|
"""
|
|
import argparse
|
|
import os
|
|
import sys
|
|
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
|
|
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
|
from src.common import load_jsonl
|
|
from nnu_common import make_raw_dataset, raw_ds
|
|
|
|
|
|
def main():
|
|
ap = argparse.ArgumentParser()
|
|
ap.add_argument("--rows", required=True, help="jsonl with {key, img, label} (native paths)")
|
|
args = ap.parse_args()
|
|
rows = load_jsonl(args.rows)
|
|
n = make_raw_dataset(rows)
|
|
print(f"[nnu:prepare] {raw_ds()}: {n} cases", flush=True)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main() |