Open CL

June 29, 2017

High-resolution visual recognition needs high-performance CNNs

Quadrupling the performance of a dedicated CNN engine within an embedded vision processing core brings more complex graph processing within reach.