Inspiration
Our main inspiration for this project were snowflakes. Snowflakes under microscope closely resemble the crystalic structure we find in dryed tears under microscope. We looked at the main techniques used for this type of research and used it in our project for predicting diseases based on images of dryed tears. Snowflakes are very structured structure which is structured structurally which makes it easily describable by methods like trees.
What it does
Our neural network is recognizing very accurately whether person has a disease or not just based on the tear data. It is also pretty good at categorizing specific diseases in relation to the amount of data provided.
How we built it
Challenges we ran into
We unfortunately couldn't find any significant distinctive metric between "SklerosisMultiplex" and "Glaukom" in some images.
Accomplishments that we're proud of
- multiplication of the data
- our metrics which showed clear destinction between some diseases
- simple cli tool
What we learned
- creating functional image classifier is a challenging task and we had to try a lot of, libraries, techniques, analysis methods we we not aware of
Log in or sign up for Devpost to join the conversation.