Covid-19 chest X-ray image classifier
The model is based on a convolutional neural network which takes a chest x-ray as an input and determines whether it is that of a healthy person's or a covid infected patient's.
Created on 16th April 2022
•
Covid-19 chest X-ray image classifier
The model is based on a convolutional neural network which takes a chest x-ray as an input and determines whether it is that of a healthy person's or a covid infected patient's.
The problem Covid-19 chest X-ray image classifier solves
The current state of the model is a base prototype which allows for single upload of images and classifies the uploaded image as covid infected or normal. This model was targetted towards the hospitals and medical centers which do not have access to modern day technology and helps in classifying x-rays which would otherwise lead to messy and scattered files which may also contain mixed up reports. Future versions of this project could include databasing the images under the patient's name under covid or normal and helps in sorting the abundant covid x-ray reports that hospitals receive. The same model can be trained under different kinds of x-ray samples such as pneumonia and a different version can be made for the same.
Challenges I ran into
Initially wanted to use an "enter a url" feature for images which would be more convenient than having images on the system and was able to solve the issue but however the model gave incorrect predictions when used with the url and file uploader at the same time, eventually removing the "enter url" feature.
Initially i was using the glob module in python to provide the pathname in the model but doing so rendered the file upload feature useless as i had to upload the file and provide the path at the same time. I went forward by removing the glob module entirely and directly passing in the image to the model , removing any problems of path unavailabilty or mismatch
Initially , the model was built only to handle jpeg images, however, streamlit's file handler was able to smoothly consider other file formats as well