- Using CNNs to diagnose pneumonia from chest X-ray images.
- Explanation of the project's aim to enhance diagnostic accuracy using advanced technology.
Description of the "Chest X-Ray Images (Pneumonia)" dataset sourced from Kaggle.
Organization of data into train, test, and validation sets with categories for Pneumonia and Normal images.
Conventional CNN model and pretrained VGG16 model. Deep learning framework used: TensorFlow.
- Architecture from input to output layers.
- Explanation of each layer's function including convolutional, max pooling, dense, and dropout layers.
*Composition of the VGG16 network.
- Adjustments made for pneumonia classification and the rationale behind using pretrained models.
- Techniques for loading, resizing, and augmenting the data.
- Importance of normalization.
- Configuration of models including optimizer and loss function selection.
- Training duration and conditions.
- Split details and methodology for training/validation/testing.
- Use of ROC curve and AUC for performance evaluation.
- Presentation of training outcomes, accuracy, and loss metrics.
- Discussion on overfitting observed during validation and the steps taken to address it.
- Summary of the study's findings.
- Assessment of the models' effectiveness in improving pneumonia diagnostics.
- Future directions for research to enhance model generalizability.
- Comprehensive listing of all sources cited in the report.
For full report, you maybe download within this repository.
Files name= 21_pages_report.pdf