The problem AI HEALTH CHECKER solves

Lack of awareness of the symptoms often lead to greater complications. People often fail to consult on time due to negligence, cost factor , time constraints, financial matters and many more. Getting an appointment from doctor takes weeks even after having all reports at hand, so one can easily input those datas in our web app and get instant AI based diagnosis of seriousness of that disease. Fake news make it more complicated to diagnose by yourself, here prediction is done based on real life dataset taken from kaggle, and the accuracy we have reached is approximately 91%(considering all diseases). Seeing your weight on scales can be frustrating, which is also a major part of diseases.
We built a web app with the help of JS, Python ,Flask framework etc, deployed on Heroku which shall tell the user chances of developing a particular medical condition or disease depending upon their symptoms using ML algorithms (Deep Learning etc). A one in all app that predicts for a wide range of diseases : Heart, Liver, Kidney, Breast Cancer, Corona, Diabetes, Pneumonia in a single place. Seeing weight can be frustrating so a little fun game to incorporate concept of space which checks your weight in different planets to lighten up the mood.

Challenges we ran into

1. Deploying all the prediction models into the web app was a huge problem we faced which was solved by using Heroku. We first uploaded all the codes in github and then deployed that using Heroku. after which we linked all the web app heroku pages in the index.html file so that it becomes easier to access directly from the home page.

2. While making the prediction models we found it difficult in few datasets to fit the trainning data and get accurate test data. This was solved after we tried and tried different deep learning approaches and studied the CNN network thoroughly.

3. Linking the models and predicting the result using flask in the "app.py" file was also a problem. There we found it difficult to pass the predicted result to the result page which was later solved after a bit of brainstorming, reading few articles and debugging.