The drive to keep getting better!
What it does
A solution for low-power, low-bandwidth, offline, and accurate AI (Computer Vision/NLP) on Raspberry Pi like device.
How we built it
We use a special compression algorithm to highly compress (up to 90%) the Computer Vision/NLP algorithms and use the computing power packed in an Nvidia Jetson Nano to run our compressed algorithms
Challenges we ran into
- Limited time to train the models
Accomplishments that we're proud of
- We have a demo of the solution we are proposing
What we learned
- Problems faced while pushing the Raspberry Pi to the limit
- Ways to deal with running in real-time large Deep Learning models on edge devices like Raspberry Pi and Nvidia Jetson Nano
What's next for The Compressor and the Compressor
- A ready-to-use, full-fledged prototype
We have a Prezi slide-deck - https://prezi.com/p/pv_vftdwv9fn/