/Covid-19-Detection-Few-Shot-Learning

Covid-19 Detection Experiments

Primary LanguageJupyter NotebookMIT LicenseMIT

Covid-19-Detection using Metrics based Few-Shot Learning

This repository is implementation of Covid-Classification models using deep learning approaches. [Paper Link] (https://doi.org/10.1117/12.2581496)

We have used following approaches for our experiments.

  1. Logistic Regression (Baseline)
  2. Convolutional Neural Networks.
  3. Transfer Learning.
  4. Siamese Networks (Few-Shot Learning)
  5. Unsupervised learning (TSNE+PCA)

Datset:

  1. https://www.kaggle.com/tawsifurrahman/covid19-radiography-database
  2. https://www.kaggle.com/pranavraikokte/covid19-image-dataset

If you found our work useful, please consider citing us.

Link: https://doi.org/10.1117/12.2581496

@inproceedings{10.1117/12.2581496,
author = {Shruti Jadon},
title = {{COVID-19 detection from scarce chest x-ray image data using few-shot deep learning approach}},
volume = {11601},
booktitle = {Medical Imaging 2021: Imaging Informatics for Healthcare, Research, and Applications},
editor = {Thomas M. Deserno and Brian J. Park},
organization = {International Society for Optics and Photonics},
publisher = {SPIE},
pages = {161 -- 170},
keywords = {Deep Learning, Classification, Few-shot Learning, Less data, COVID-19},
year = {2021},
doi = {10.1117/12.2581496},
URL = {https://doi.org/10.1117/12.2581496}
}

Embeddings Visualization of Covid-19 CT scan dataset

alt text