- This is an implementation of a course project conducted in June 2020 for the postgraduate course EC736-Neural Networks Applications. The project resulted in a research paper, which can be found at: COV-NET_COVID19_Detection_Using_Neural_Network.pdf
- Code Implementation for binary classification (COVID-19 and pneumonia): Covid_2Classes.ipynb
- Code Implementation for multi-class classification, (COVID-19, pneumonia, and normal): Covid_3Classes.ipynb
- Additionally, A code implementation to the kaggle dataset is located at: Pneumonia-KaggleDataset.ipynb
Dataset used in this repo can be accessed through this link
A new chest X-ray dataset for COVID-19 detection was created by combining two existing datasets. First dataset from kaggle consists of pneumonia and normal cases, while the second dataset from Github exclusively contains COVID-19 cases. The constructed dataset facilitates the evaluation of models for two tasks: binary classification (COVID-19 and pneumonia) and multi-class classification, (COVID-19, pneumonia, and normal).
| Model | Binary-class Accuracy | Multi-class Accuracy |
|---|---|---|
| ResNet18 | 97.64% | 85.71% |
| Mobilenet_v2 | 98.43% | 82.88% |
| Shufflenet_v2 | 99.02% | 85.04% |
| Shufflenet+Mobilenet | 99.80% | 84.77% |
| Resnet18+Shufflenet | 99.41% | 88.01% |