There is a newer version of the record available.

Published May 20, 2026 | Version 1.0

Chest X-Ray Dataset for Pulmonary Disease Classification (COVID-19, Tuberculosis, Pneumonia, Normal)

  • 1. ROR icon Universidade Federal de Santa Catarina

Description

Dataset Overview This dataset contains chest X-ray (CXR) images organized for the classification of four pulmonary conditions: COVID-19, Tuberculosis, Pneumonia, and Normal. It was compiled from multiple publicly available sources and is intended for training and evaluating Deep Learning models for medical image classification.

Classes (4 Labels):

  1. COVID-19;
  2. Tuberculosis;
  3. Pneumonia;
  4. Normal;

Dataset Split Strategy The dataset is pre-split into three subsets following a 70/15/15 ratio, ensuring a consistent and reproducible evaluation benchmark:

Class Total Train (70%) Val (15%) Test (15%)
COVID-19 2,300 1,610 345 345
Tuberculosis 2,300 1,610 345 345
Pneumonia 2,300 1,610 345 345
Normal 7,100 4,970 1,065 1,065
Total 13,900 9,800 2,100 2,100

Strict separation was maintained between subsets to prevent data leakage.

The script used to aggregate and split the dataset is available as build_dataset.py

Sources & Licenses This dataset aggregates images from the following publicly available sources:

Source Used for License
Tuberculosis Chest X-rays Images — Kiran & Jabeen, Mendeley Data, 2024 Tuberculosis CC BY 4.0
Labeled Chest X-Ray Images — Kermany et al., Mendeley Data, 2018 Pneumonia, Normal CC BY 4.0
COVID-19 Chest X-Ray Image Repository — Figshare COVID-19 CC BY 4.0
COVID-19 Image Repository — ml-workgroup COVID-19 CC BY 3.0
Radiopaedia (via ieee8023/covid-chestxray-dataset) COVID-19 CC BY-NC-SA
Eurorad (via ieee8023/covid-chestxray-dataset) COVID-19 CC BY-NC-SA
Coronacases (via ieee8023/covid-chestxray-dataset) COVID-19 Apache 2.0


The combined dataset is released under CC BY-NC-SA 4.0, reflecting the most restrictive license among the contributing sources.

If you use this dataset in your research, please cite it as in Citation

Files

dataset.zip

Files (1.3 GB)

Name Size
md5:8245f9d6f8935375b4d08e43a1d7c503
10.2 kB Download
md5:6e82f30a8f02ce450a1e725b8e57855b
1.3 GB Preview Download