An image classifier to classify Chest X-Rays using Transfer Learning.
Chest X-Rays are notoriously difficult to interpret when starting out for any new doctor. By using an classifier, we can help with the triage process to risk stratify x-rays into high suspicion of pneumonia, or low suspicion.
This model was trained using this dataset, using Transfer Learning to retrain the last layer of a MobileNet model.
| Class | Accuracy | # Samples |
|---|---|---|
| Normal | 0.94 | 202 |
| Pneumonia | 0.98 | 582 |
The model is implemented using Tensorflow.js, which enables local classification rather than sending data to a server.
Live demo: www.harvinder.me/sea-eagle
- Allow users to select from a list of example Chest X-Rays for testing
- Drag and drop functionality