Pneumonia Detection using Transfer Learning

Approach

Pneumonia detection is commonly predicted using basic CNN models. However , the accuracy of these state of the art models is pretty low. On the other hand, our approach of using Transfer learning, that is, by using VGG-19 and ResNet50, we can get a much better accuracy.

Project

Our project consists of 3 approaches:

• VGG-19 Model
• ResNet50 Model
• ResNet50 Model after Fine-Tuning parameters

Dataset

Dataset consists of images of 2 classes of chest X-ray images:

1. Diseased
2. Normal

                   

The dataset can be accessed from the Data Folder.

Model Results

VGG-19

         

ResNet50

         

ResNet50 - Fine Tuned

         

How to Use

• Clone the repository using :

    $ git clone https://github.com/rishusiva/Pneumonia_Detection
  

• Enter the directory using:

    $ cd Pneumonia_Detection/
  

• Install the requirements using:

    $ pip install -r requirements.txt
  

• Run the demo notebook

Results

• VGG-19 : 85%
• ResNet50 : 91%
• ResNet50 Tuned : 95%

Project Maintainer(s)

Rishikesh S

Ananya Negi

Contribution

Contributions are always welcome! You can contribute to this project in the following way:

• Increasing the accuracy
• Bug fixes if any
• Creating an application

Do check out the documentation for Contribution Guidelines.