Detecting COVID-19 in X-ray images with Keras
This repository uses Adrian Rosebrock's bog as reference.
Disclaimer
Due to very limited dataset(25 positive images) the model is not meant to be a reliable COVID-19 diagnosis system, nor has it been professionally or academically vetted.
Dataset
- 25 Positive images were obtained from the following rep
- 25 negative images (healthy patients) were obtained from Kaggle’s Chest X-Ray Images (Pneumonia) dataset
So now we have a balanced dataset. But the number of images are still very less. So data augmentation was applied. (refer to the colab notebook)
Architecture
Fine-Tuned VGG16 architecture.
Directory structure
├── Dataset
│ ├── covid [25 entries]
│ └── normal [25 entries]
├── Covid-19.ipynb
├── Results
├── model.png
Results and conclusion
precision recall f1-score support
covid 0.83 1.00 0.91 5
normal 1.00 0.80 0.89 5
accuracy 0.90 10
macro avg 0.92 0.90 0.90 10
weighted avg 0.92 0.90 0.90 10
- Accuracy - 90%
- Recall(True Positive Rate) - 1. This means our model correctly classifies all the positive cases.
- Specificity (True negative rate)- 0.8 This means only 80% were correctly identified as healthy out of all the healthy patients.