Pneumonia RX Classifier is a CNN model created with TensorFlow in order to distinguish type of pneumonia from RX image.
Colab Work is aviable at:
https://colab.research.google.com/drive/1qAAclorvsUkmgDXpHhUG267Ibh3UkblL?usp=sharing.
The dataset used in this project is Chest X-Ray Images (Pneumonia), aviable at:
https://www.kaggle.com/paultimothymooney/chest-xray-pneumonia.
It contains 5,863 images of RX.
Dataset is splitted in three parts:
- Training
- Validation
- Test
In this project I used only training and validation parts.
The classification labels are:
- NORMAL : RX without pneumonia
- PNEUMONIA_BACTERIA: RX with Bacteria Pneumonia
- PNEUMONIA_VIRUS: RX with Virus Pneumonia
For my purpose I used a CNN model with the following architecture:
The results with batch size 32 and 6 epochs are the following.
| model | accuracy |
|---|---|
| RX Pneumonia Detector v1.0 | 62,34% |
- Improve model performances
- Measure system with precision, recall and f1 measure
- Data Augmentation
Samir Salman