New COVIDx8B dataset is created in a way where files like COVID(1).png or COVID(2).png are in the labels file, but not in the actual data.

[chododom]

It seems that the real name of file COVID(1).png is COVID-1.png and analogously for the others.

[chododom]

I have solved this by renaming the files in my scripts followingly:

labels_file = open(labels_file_path, 'r')
while True:
    line = labels_file.readLine()
    if not line:
        break
    split = line.split(' ')
    if split[-3].startswith('COVID(')
        name = split[-3].replace('(', '-')
        corrected_filename = name.replace(')', '')

All other files seem to be present.

[mayaliliya]

Hi @chododom, the COVIDx8 dataset was created with Version 3 of the COVID-19 Radiography Database which can be downloaded from here: https://www.kaggle.com/tawsifurrahman/covid19-radiography-database/version/3

Please use this version of the dataset for COVIDx8 as the images do not map directly between Versions 3 and 4. We will make note of this versioning in the COVIDx docs moving forward, thank you for bringing it up!

[chododom]

Ohh okay, thank you @mayaliliya , maybe it would be good to include this in the guide to setting up the dataset, cloning the repo gets the latest version :)

[khanbhai0078]

In what place you have added this code?