Root Cause

• 

  Root Cause

Inspiration

Whenever our mothers try to take care of plants, they always end up dying. This led us to come up with the idea of an app that helps you keep your plants healthy.

What it does

The web app gets the video stream from your device's camera, and allows you to take a picture of your plant. This picture is sent to a custom-made API, where the plant is identified using a neural network, and it is determined if it has any diseases. The user receives the neural network's top three predictions, selects the one that is right for them, and can ask an advanced LLM for advice on how to care for their plant and help it recover from its ailments. This request is once again processed by our custom API and returned to the user.

How we built it

The custom-made API is hosted on a Lambda GPU instance running Ubuntu. PyTorch was used for training the neural network responsible for plant classification as well as running the model to process requests. The Lambda inference API was used to run LLMs and pass in the necessary data to give the user a relevant response. The API itself was created using FastAPI.

Challenges we ran into

Hosting the site publicly wasn't possible due to the limitations of communication between a HTTPS frontend and our HTTP backend. Unfortunately, after around an hour of trying, we realized that securing the backend API properly wasn't feasible for the project.

Accomplishments that we're proud of

We learned how to use many new tools, and create a relatively robust application. Creating an API and being able to send requests to it was a great accomplishment. Surprisingly, during the hackathon, we managed to stay focused for a majority of the time.

What we learned

Custom domains are great, but they introduce numerous challenges that even users with experience in the topic will have trouble getting them to work properly.

What's next for Root Cause

Getting the API to use HTTPS would be the main next step, but would likely involve changing service providers, which could get complicated. Enlarging the dataset for plant classification would also be a good expansion. Some sort of feature where users interact with each other and exchange ideas would be a long-term goal.

Built With

• api
• javascript
• lambda.ai
• python
• pytorch
• ubuntu
• vite