July 12, 2021
A paper titled “Hardware-aware Model Optimization Tool For Embedded Devices”, authored by Cagri Ozcinar, Dongsung Kim, Benjamin Rufus Duckworth, Shayan Joya, Nicolas Scotto Di Perto, Attila Dusnoki, Márkó Fabó, Dániel Vince, Gábor Lóki, Ákos Kiss, and Christopher Alder has been accepted for and presented at the ICME conference.
In the article, which was first presented at this year’s IEEE International Conference on Multimedia and Expo (ICME), the authors propose a new hardware-aware model optimization tool, called HOT, which will find optimal operators for deep neural network models based on the given target-hardware, thereby improving the latency and accuracy of the models.
The paper can be read here.
Page last modified: July 12, 2021