This repository contains code for segmentation and localization tasks in Computer Aided Detection for Pulmonary Embolism (CAD-PE). The project uses a UNet model for segmentation and a Faster R-CNN model for localization. The dataset is structured into separate folders for localization and segmentation tasks.
-
cadpe_localization/
:data/
: Localization dataset
-
cadpe_segmentation/
:dataset/
: Segmentation datasetmodel_org/
: Segmentation modeltrain/
: Segmentation training pipeline
-
images/
: Visualization images for CAD-PE segmentation and localization -
notebooks/
:- Notebooks for both localization and segmentation tasks
-
utils/
:coco_format.py
: Utility script to generate COCO formatcoco_dataset.json
: Dataset JSON file in COCO formatgetbbox_and_plot.py
: Utility script for getting bounding boxes and generating plots
-
Clone the Repository:
git clone https://github.com/your-username/CAD-PE-SegLoc.git cd CAD-PE-SegLoc
-
Data Preparation:
- For localization, place your localization dataset in
cadpe_localization/data/
. - For segmentation, place your segmentation dataset in
cadpe_segmentation/dataset/
.
- For localization, place your localization dataset in
-
Training:
- Navigate to
cadpe_localization/train/
for localization model training. - Navigate to
cadpe_segmentation/train/
for segmentation model training.
- Navigate to
-
Notebooks:
- Explore the
notebooks/
directory for Jupyter notebooks covering analysis, training, and evaluation.
- Explore the
coco_format.py
: Use this script to convert your dataset to COCO format.coco_dataset.json
: COCO format dataset JSON file.getbbox_and_plot.py
: Script for extracting bounding boxes and generating plots.
-
This project utilizes the PyTorch implementation, developed by the PyTorch community. Check out their official repository: PyTorch.
-
If you use or reference the CAD-PE dataset, please provide proper attribution as per the following publication:
@article{gonzalez2020computer, title={Computer aided detection for pulmonary embolism challenge (CAD-PE)}, author={Gonz{\'a}lez, Germ{\'a}n and Jimenez-Carretero, Daniel and Rodr{\'\i}guez-L{\'o}pez, Sara and Cano-Espinosa, Carlos and Cazorla, Miguel and Agarwal, Tanya and Agarwal, Vinit and Tajbakhsh, Nima and Gotway, Michael B and Liang, Jianming and others}, journal={arXiv preprint arXiv:2003.13440}, year={2020} }
Feel free to customize the content based on your specific project details and requirements.