NeMo-Net

Inspiration

We wanted to use the reasoning power of NVIDIA's Nemotron AI to analyze data from and control multiple devices simultaneously.

What it does

The goal was to have a network with multiple Raspberry Pis with cameras and various sensors which would relay their data to a central control hub where Nemotron would be able to analyze the data and make decisions on what actions the devices should take. Unfortunately we were not able to realize this vision in time, leaving the final product with little functionality.

How we built it

On the Raspberry Pi we had a python script which sent images from a webcam to a central hub program running on a different computer over websockets. The central hub program was then supposed to collect multiple images and send them to Nemotron through it's API for it to analyze.

Challenges we ran into

We all had fairly little experience with software development as a whole, so we had to spend a lot of our time researching the tools to make it possible, such as networking systems and using the Nemotron API. We also struggled bringing everything together since we ended up each developing a different part of the system mostly independently of each other.

Accomplishments that we're proud of

We were able to get the websockets working for sending images from the Raspberry Pi to the central hub program.

What we learned

We learned about many foundational aspects of software development such as API calls and networking.

What's next for NeMo-Net

We want to get the base system for Nemotron analyzing the images being sent from the pi to work, and then possible continue developing it to allow for decision making based on these analyses.