PneumoniaNetResnet50

Description

Main objective of this program is to classify Pneumonia with deep learning methods. Through this model i hope it can help health workers to automate process classifying Pneumonia throug chest ray image

My main method to make this is using a deep learning approach. I use Convolution Neural Nets to create the classifier and apply transfer learning with pre-trained models. I used ResNet50 as the backbone fine-tune it with my own dataset (which I got from this Kaggle link). For the architecture, I used Pytorch as the base framework.

Installation Guide

Docker

Build the Image ````DOCKER_BUILDKIT=1 docker build -t clayrisee/pneumonianet:resnet50 . ```

Run Container ````sh runDocker.sh```

Github

• Clone this repository in your local with this command git clone https://github.com/Clayrisee/PneumoniaNetResnet50.git
• Install the requirement modules pip install -r requirements.txt
• If the module already satisfied, you can run the inference.py

Main.py

To use main.py, the model is in .onnx format. My .onnx saved model is in this folder.

1. See details arguments python main.py -i

2. Perform inference python inference.py -i <your_image_path>

Result

Input Image	Output
	Pneumonia Confidence:99.98%
	Pneumonia Confidence: 99.83%
	Pneumonia Confidence: 97.71%

Othe Version using ResNet50

I Also build another version of this project using Vgg-19 backbone, you can visit this Repository