Detecting Pneumonia with Convolutional Neural Networks
Main Code: Pneumonia_Diagnosis.ipynb
Detecting Pneumonia Diagnosis on a small and challenging dataset that contains Xray iamges of Patients with three different Convolutional models.
1. Fine tuning.
Training all the base-layers of VGG-16 once again + the newly added fully connected layers for better accuracy.
Results: 1. Val_accuracy: 0.8750 2.Evaluate on Test data: 0.9359
2. Transfer learning.
Training only the newly added fully connected layers of our network.
Results: 1. Val_accuracy: 0.9375 2. Evaluate on Test data: 0.9295
3. Depthwise Convolutional Network.
Training with our own depthwise Convolutional Network
Results: 1. Val_accuracy: 0.8750 2. Evaluate on Test data: 0.8958
The famous VGG-16 network that I used for first two models for finetuning and transfer learning respectively.
In the below image, the network on the right is an example of the fine-tuningka.
