How to Make Bad Deep Learning Hardware
We have a new style guide with hints about writing and typesetting.
The Imaging Pipeline
Mark’s full ISCA 2018 talk about accelerating live vision is now on YouTube.
Please Help Me Solve an Old Approximate Computing Problem
Watch Mark’s lightning talk for our ISCA paper on redundancy in real-time video.
Program Synthesis is Possible
We’re collecting raw photos for a new computer vision dataset.
If you have an Android device, you can help out by downloading our app.
We’re looking for a postdoc to help us build a new kind of continuously reconfigurable machine (and a compiler to go with it).
Our paper on exploiting temporal redundancy in hardware-accelerated computer vision was accepted to ISCA 2018.
All About Invention Disclosures
Patenting in Academia and Industry
Our paper on programming language support for NLU will appear at SysML 2018.
FODLAM, a Poorly Named Tool for Estimating the Power and Performance of Deep Learning Accelerators
Our paper about Braid appeared at OOPSLA 2017. There’s a video of the talk on YouTube.
Mark and Suren presented our poster about our work on efficient cameras for computer computer vision at ICCV 2017.
Closed Problems in Approximate Computing