A data preparation and model building notebook on pneumonia classification, created on kaggle
Dataset available here
The normal chest X-ray (left panel) depicts clear lungs without any areas of abnormal opacification in the image. Bacterial pneumonia (middle) typically exhibits a focal lobar consolidation, in this case in the right upper lobe (white arrows), whereas viral pneumonia (right) manifests with a more diffuse interstitial pattern in both lungs. The dataset is organized into 3 folders (train, test, val) and contains subfolders for each image category (Pneumonia/Normal). There are 5,863 X-Ray images (JPEG) and 2 categories (Pneumonia/Normal).
Chest X-ray images (anterior-posterior) were selected from retrospective cohorts of pediatric patients of one to five years old from Guangzhou Women and Children’s Medical Center, Guangzhou. All chest X-ray imaging was performed as part of patients’ routine clinical care.
Data is imabalanced with 3000+ pneumonia images and 1000+ normal images
These images look almost similar no matter the case and therefore feature extraction would be quite difficult, I tried to acheive a precision of 0.93 despite of maximum images being of pneumonia. Issue a pull request if acheived better results :)
