Study on transfer learning capabilities for pneumonia classification in chest-x-rays images
This repository contains the official code of the research paper Study on transfer learning capabilities for pneumonia classification in chest-x-rays images pubblished at the Computer Methods and Programs in Biomedicine Journal.
Cite this work
@article{avola2022study,
title={Study on Transfer Learning Capabilities for Pneumonia Classification in Chest-X-Rays Images},
author={Avola, Danilo and Bacciu, Andrea and Cinque, Luigi and Fagioli, Alessio and Marini, Marco Raoul and Taiello, Riccardo},
journal={Computer Methods and Programs in Biomedicine},
pages={106833},
year={2022},
publisher={Elsevier}
}How to install
You can download a copy of all the files in this repository by cloning the git repository:
git clone https://github.com/andreabac3/study-transfer-learning-covid-19Then you should install all the dependencies. To do this we suggest to use the setup.sh file to create a new conda environment.
bash setup.shDataset construction
To be compliant to copyright issue, we release the file list that compose our dataset.
To build the dataset:
(i) You should download the Kaggle data from the competition page.
(ii) For the covid data you can download the images from this github repository.
(iii) The downloaded data must be re-arranged following the dataset structure present in the file data/mia/dataset/dataset_structure.txt .
mia
└── dataset
├── test
│ ├── BACTERIA
│ │ ├── person100_bacteria_475.jpeg
│ │ ├── person100_bacteria_477.jpeg
...
To run your experiments
You can use the train_and_test.sh file.
bash train_and_test.sh