Fish King

Inspiration

We wanted to design something fun that would bring together a community that doesn't have as many ways to come together that are tailored to what they want out of their hobby! We thought of apps like strava as our example of how to gamify and create community around fishing.

What it does

For each user, you can upload photos of catches to the app or website. It can use the photo's metadata to classify where the image is and it uses a computer vision model to classify species and size! You can compete with fishermen in the same lake on a leaderboard to see who can catch the biggest fish in the lake and become the fish king. We also have an ensemble of machine learning models that help fishers know what parts of the lake to catch certain fish at, and how good certain days are for their odds of landing a big catch for each and every species they'd want to find. Finally, we have a badge system to give fishers another way to keep track of their accomplishments and progress.

How we built it

For the app we used mostly swift. For the backend we develop a backend using SQLite. For the database they are several of themn and use python script and LM Studio to load and run the model. For the model classification we used a Qwen 2.5 7b-instruct{computer vision}. For ML we used datrobot to find the best one.

Challenges we ran into

We were having problem integrating all piece and making sure the website and the app was well connected. Also model classification needed a bit of help.

Accomplishments that we're proud of

Working ML model ensemble with realtime weather API data. Used MLops State of the Art process to fine tune and iterate our models. Local-hosted a Qwen based computer vision model to classify fish and size. connected frontend and backend, working leaderboard, web and app interface.