FishFinder

Unfinished design

Inspiration

We wanted to create something that taught us machine learning, a field we've never seen before, and create something that could have a purpose.

What it does

Scans fish through your phone's camera or through uploading on a website, and identifies what kind of fish was given.

How we built it

We used the TensorFlow API as well as Google's image recognition software

Challenges we ran into

Using number data, such as length and weight was too generic
Creating a website that can communicate with python scripts was exceptionally challenging, as our only option for hosting was Microsoft Azure, which did not support installing external python packages

Accomplishments that we're proud of

Creating a simple, yet effective machine learning program
Though we weren't successful with it, getting comfortable with the daunting Azure workspace was very rewarding
Understanding and being able to explain how machine learning, and a lot of day-day apps work (i.e. YouTube giving recommendations)
Working with PHP and Django (though they weren't successful)

What we learned

What is machine learning?
How do you implement machine learning into many different applications?
How can we run Python scripts through a web app?

What's next for FishFinder

Implementing it as a web-based application, or even an Android/iOS app. We could also manually expand our database in order to increase the accuracy of our program, or possibly find much larger databases online.