AquaVison

Inspiration

We were inspired by the on going struggle of not having enough water especially with the ongoing development of AI. As technology improves, it's our responsibility to find proper ways to take care of our planet for future civilians so they have a place to call home.

What it does

AquaVision is able to detect rooftops that have more than 100,000 square feet and contains a cooling tower on top of it using it's AI detection. Additionally it is able to give a viability score for each qualifying building and show reports of how much water is saved if Grundfos products are implemented.

How we built it

For this project, we used FastAPI to handle to API calls for the backend and YOLO for training the AI with CVAT to annotate and collect the samples. We also used HTML, CSS, and JS for our frontend and Python for our backend.

Challenges we ran into

Our biggest challenge was training the AI as cooling towers were very hard to distinguish especially with HVACS as they look very similar. Additionally, when using CVAT, when we exported the annotations, it didn't export all of them and from the ones it did, some were corrupted. This made it difficult to train the AI properly with good samples.

Accomplishments that we're proud of

We are proud that we were able to use API Calls to flag and calculate the square feet of roof tops and detect cooling towers so the user can know what buildings would be more viable to sell Grundfos products to in order to save and reuse water.

What we learned

From this experience we learned how difficult it is to train AI and how to distinguish between good and bad samples.

What's next for AquaVison

For the future, we plan to get more samples to enhance our AI detection to give more accurate results. Additionally, we plan to make a search feature to making finding locations more feasible and create a login that can save bookmarked buildings that you plan to visit to see if they're willing to collaborate with Grundfos to save water for the future.

Built With

• css
• fastapi
• google-maps-javascript-api
• html
• javascript
• python
• uvicorn