Chest_XRay_Classification

General Instructions

Step 1 - clone this Chest_XRay_Classification GitHub folder

• In Github, create a personal token if not yet done so (Settings -> Developer Settings -> Personal access tokens). This will be used when logging in to GitHub through the git clone command.
• Clone this Chest_XRay_Classification GitHub folder on local device, using

git clone https://github.com/YutingGu/Chest_XRay_Classification.git

• You will be asked to enter the password to log in, which is your personal token.
• After logging in, the GitHub folder will be successfully clone to your local device.

Step 2 - Image Data Download

The data used for this project includes image data and label data. Since image data have not been uploaded to GitHub, we need a further step to download the image data to local device.

• Run all cells in Dataset_Download.ipynb to download and unzip the image dataset.

Note:

• Dataset will be downloaded under the the folder Dataset/images/. All files end with .tar.gz. can be deleted after running all cells to save space, these are the compressed image files.
• Please make sure in your .gitignore file, Dataset/images/ is included to avoid uploading the whole dataset

The 'Dataset' folder

This folder contains the original data downloaded from https://paperswithcode.com/dataset/chestx-ray8, and the processed data (one-hot encoded labels for train/validation/test set) after running Utilities/label_generator.py.

The original data:

• CXR8_Data_Entry_2017.csv, test_list.txt, train_val_list.txt

Processed data:

• train_list.txt, val_list.txt,
• test_label.csv, train_label.csv, val_label.csv

The following is the method for doing one-hot encoded label generation and dataset split:

• Change directory to Utilities, use terminal to run the label_generator.py to generate one-hot encoded labels for train/validation/test set.