CheXpert classification

Purpose

This repository attempts to classifying 5 classes of CheXpert dataset (Cardiomegaly, Endema, Atelectasis, Pleural Effusion) using different combination of training techniques.

Dataset

CheXpert is a extensive collection of chest X-rays used for automating chest X-ray analysis. This dataset includes uncertainty labels and sets evaluated by radiologists, serving as a benchmark for automated interpretations. To download the dataset, visit this kaggle website: https://www.kaggle.com/datasets/mimsadiislam/chexpert

Installation

Clone the repository

git clone https://github.com/nguyenbanhducA1K51/chexpert_classification.git

cd to the subdirectory

cd chexpert_classification/chexpert

and use the package manager pip to install libary and package in file requirements.txt (recommend install in conda environment).

pip install -r requirements.txt

Usage

• Open the ../chexpert/config/config.yaml file and change the value of following variable: "project_path": "path where you clone the this repository" "data_path": "path where you download your dataset" " process_train": "path that will save train/val csv file after process them" "process_test": "path that will save train/test csv file after process them"

• First , run the preprocess file at "..chexpert/preprocess.py"

python3 preprocess.py

• Run the training pipeline by

bash chexpert/scripts/train.sh