Clarity Lab over at the University of Michigan published a really interesting paper: Sirius: An Open End-to-End Voice and Vision Personal Assistant and Its Implications for Future Warehouse Scale Computers. Clarity Lab also has a website for the project: Sirius: An Open Intelligent Personal Assistant. Sirius is open source.
Why mention this on the Jetson blog? Because the technologies used in this large ecosystem are constructed from bits and pieces which we’ve discussed here before. OpenCV is used for Image Matching. Caffe is used for Deep Neural Network scoring in Speech Recognition. The System runs on Ubuntu 14.04, which is equivalent to Linux for Tegra on the Jetson. In these large data centers, computational power efficiency is a major key to the game. We know that, while offering great performance, traditional GPUs can be a little piggy when it comes to power. Not so with the Tegra K1 (or the upcoming Tegra X1).
This sounds like a very interesting area. Certainly the paper and website are worth spending some time reading and looking around even if you’re just curious as to how Apple’s Siri, Google’s Google Now or Microsoft’s Cortana work on the inside. Magic? Maybe not, but it’s certainly impressive.
One Response
I know this article is quite aged. However i would like to inform you that lucida the updated version of sirius works on the nvidia jetson tx2. Only a few things have to be modified ie: (-msse4.2) flags in folly and it will work perfectly. You should make a demo on it !