Gratheon

Monitor bees with AI

Comment

Inspiration

Beekeepers lose colonies due to varroa mites, hornets, late swarming. They need to often inspect the hive which is a hard physical labor. We want to help. Full list of problems: https://www.notion.so/gratheon/Problems-we-are-solving-15a899e8bf10455c9ef903c6e269af2c?pvs=4

What it does

Monitoring bees with AI. We help beekeepers with observability, so that they could be more efficient and their colonies would stay healthy. We provide app and integrate hardware so that computer vision could alert a beekeeper in case of danger. Our long-term vision is autonomous robotic beehive.

For the hackaton, we plan to implement some of the features:

  • LLM beekeeping prompt. Basically, given a state of the beehive, LLM should give advice what to do
  • Integrate queen or varroa mite detection based on image input. Likely needs testing of existing models or training it from own dataset.
  • Integrate bee counting inferencing model. Given a 10 sec video file, provide a count of bees entering/leaving a beehive.

Full list of features that we can work on -> https://www.notion.so/gratheon/Feature-hub-ca76f493a1454e14b7dfbae7bd292a2e?pvs=4

How we built it

Current web-app is capable of receiving a photo of a beehive frame, detecting bees, queen cups and frame cell types. A prototype can also count bees at the hive entrance, based on video input.

We use Apollo's Federated GraphQL under the hood and urql client in the web-app, which is also PWA. Mobile app for google play is in testing phase. https://gratheon.notion.site/Gratheon-fdac3638d55542ddb073ae90537cd5b5

We use Clarifai to detect queen cups. A disclamer, founder also works in Clarifai.

Challenges we ran into

We need more AI models, more real-life integration with edge devices, inferencing on mobile phones.

Accomplishments that we're proud of

Web-app at its current state that is fully opensource, you are welcome to contribute https://github.com/gratheon

What we learned

AI model inference needs dedicated hardware with GPU. Optimizing models is hard.

What's next for Gratheon

We need a strong team, raise capital, build SaaS platform for monitoring https://www.notion.so/gratheon/Feature-hub-ca76f493a1454e14b7dfbae7bd292a2e?pvs=4

And two physical devices: https://www.notion.so/gratheon/Hive-entrance-bee-counter-fd9559a2950b44bc8291972299ced18e?pvs=4 https://www.notion.so/gratheon/Robotic-nest-b013573cc23940fca32748dfe6b0210a?pvs=4

Individual tasks for app are available at https://github.com/orgs/Gratheon/projects/1/views/1

Share this project:

Updates