X-rays images classification

• online app : https://chestxrays.herokuapp.com
• local app : if you want app to your own matchine follow the next step

Local app

1.clone repository

• click to Code > Download ZIP (then extract the folder)

2.install

• python ( if you don't have ) : https://www.python.org/downloads/
• requirements.txt : open CMD in the folder then : python -m pip install -r requirements.txt

3.using

• put the images to "predicts_custom" folder
• open CMD in the folder(make sure in folder have predicts.py) : python predicts.py
• the result in "save_dataframe" folder is a .csv file format

Under the Hood

X-rays image classifications from : Kaggle dataset

model architecture : Convolutional Neural Network (CNN)

for example CNN-architecture :

image source : https://www.theclickreader.com/building-a-convolutional-neural-network/

• has 2 conv2D layers and max-pool layers between
• 1st conv2D 16 filters, 2nd has 32 filters
• with RELU activation.
• finally flatten then 2 output layers for binary classifications.

1st conv2D 16 filters :

1st conv2D 16 filters outputs :

2nd conv2D 32 filters outputs :

some of the filters can detect the Ribs or Space between the Lung.