Download data from: nihcc or kaggle

All images from sub directory of images are copied into all_images/ folder with copy_images.py.

change csv file name from Data_Entry_2017_v2020.csv to data.csv

make a simpler data2.csv with simplify_csv.py: remove columns, rename columns, ignore "No Finding" in "Finding Labels"

<!-- original columns -->
Image Index,Finding Labels,Follow-up #,Patient ID,Patient Age,Patient Gender,View Position,OriginalImage[Width,Height],OriginalImagePixelSpacing[x,y],

<!-- columns removed -->
Image Index,Finding Labels,Patient Age,Patient Gender,View Position

<!-- renamed -->
name,label,age,gender,position

<!-- image names changed from png to jpg -->

make db.sqlite3 form data2.csv with csv_to_sqlite.py

1024x1024 png images from all_images/ that exists in data2.csv ignoring the extension of the image files are resized to 512x512 jpg images with resize_images.py into resized_images/.

⚠︎ this takes very long time e.g. 783s