Trova

• 

  Init

• 

  Enter name

• 

  pick category

• 

  object found and detect success

• 

  middle screen

• 

  leader boards

Inspiration

Caffein bars.

What it does

It's a simple object hunting game

How we built it

Using Xcode, GCP (GKE), Node.JS, and MongoDB

Challenges we ran into

Solace API documentation was hard to get around Weird GCP GKE container deployment issues Training our own models resulted in some weird results when classifying objects

Accomplishments that we're proud of

Performant live image recognition of day to day items UI/UX of our iOS apps

What we learned

Developing distributed systems that with many Microservices has its own challenges Solace was a very difficult service to use, and we encountered many difficulties with it How to properly train ML models

What's next for Trova

We were unable to add event subscriptions and publishing to our mobile app. In the future Trova should perhaps pursue other messaging services such as Kafka for developing its event driven capabilities.

We trained our own models for recognizing specific objects for our game. This is not a very maintainable workflow and Trova should try and pursue pre-trained ML models

Built With

• express.js
• gcp
• gke
• google-cloud
• mongodb
• node.js
• xcode