The images are in PNG format. The resolution of the images are 512x512.
The dataset has been divided into train,test and val set.
The lung images dataset is a comprehensive collection of images used for the training and evaluation of deep-learning models for the diagnosis of lung infections. The dataset contains a total of 17,275 images, consisting of 10,406 normal images, 5,775 pneumonia images, and 1,094 tuberculosis infected images. The images were sourced from multiple locations, including RSNA, Montgomery County chest X-ray set, Shenzhen chest X-ray, Qatar University, Doha, Qatar, and the University of Dhaka, Bangladesh, and their collaborators. The original images were in the DCM format, and they were converted to the Png format to ensure compatibility with deep learning models. This dataset is an essential resource for researchers, clinicians, and data scientists working on lung infection diagnosis, and it provides a valuable tool for the development of advanced AI models for lung disease diagnosis.
I’ve built a model with a 104-layer architecture based on residual networks, inspired by the well-known ResNet model. The model was trained on a GPU and achieves an accuracy of 80%. If I weren’t limited by GPU resources, I could make slight adjustments in the learning rate and other hyperparameters to achieve higher accuracy.
E-mail: as.alirezasaharkhiz@gmail.com
github: https://github.com/alirezasaharkhiz9
