Inspiration

Inspired by Corona virus, looked into solutions for diseases that could utilize the algorithms we knew We knew that we could implement machine learning and had some experience with html, so we decided on making a tensorflow model that would look at cells

What it does

There is a website that runs on a local server that has a home page with further description of the project, a meme page, a credits page for the images and data set used, and the page for the actual testing. Once a png is uploaded to the page, it is saved into a folder which the model reads and returns a percent chance of infection as well as a link for further steps if needed

How I built it

two teams: one that worked on machine learning algorithm and other that worked on website creation and integration The model was made in python and trained on 20,000 images and left a few thousand for testing purposes. It is run on a constant loop in Jupyter notebook and prints the percent in a json file. The website was made in php so logical operators could be used with html. The test page was made first, then more pages were added to make the full website.

Challenges I ran into

many bugs with image conversion and errors regarding developing good model architecture trouble with the integration of python with the html in the use of shared folders and json file originally planned to use jsp to integrate java within html because most of our team came in only knowing java had to scrap that code and completely learn php and css in the day from scratch many bugs with the integration of php and css within html

Accomplishments that I'm proud of

learning 3 languages in 1 day adapted a machine learning model many times to find best accuracy made a website from html after learning it completely the day before good memes program works better than expected and can be generalized for other uses debugging everything just in time managed to finish

What I learned

don't reinvent the wheel testing a lot is necessary the first solution is probably not going to be the best solution more time has to be allocated at the end for integrating parts than planned code is a lot more complicated that first envisioned it's always better to keep things simple

What's next for Blood Cell Malaria Tester

improve website layout possibility for integrating in third world countries where disease is present adding data sets for other diseases to train on method is generalized, so it could be easily adapted to fit many other use cases

Built With

Share this project:
×

Updates