The dataset contains 20,000+ images and 3 classes: COVID-19, Pneumonia and Normal (healthy).

The data are divided into train and validation sets which can be used to develop and tune the detection/classification algorithms.

A test set will be released later with a set of images to be classified into one of the 3 classes. An example submission file will  be provided. The participants will need to upload their classification results on the test set using the same format as the example submission file.

Training and validation sets

The train and validation sets contain:

  • train = 17,955 CXR images
  • validation = 3,430 CXR images
  • The use of pretrained weights is only permitted if the weights are not from other CXR COVID-19 datasets or Chest X-Ray datasets (Pneumonia,  Normal, etc.) (e.g. The use of ImageNet pretraining is permitted)

Test set

Will be available October 1st, 2021 with a submission sample file (submission file will be mainly 2 columns, the 1st column with the image names and the 2nd column with the predictions (class number): 0, 1 or 2)