Title :
A new high-speed real-time video processing platform
Author :
Muller, Johannes ; Muller, Johannes ; Tetzlaff, Ronald
Author_Institution :
Inst. of Fundamentals of Electr. Eng., Tech. Univ. Dresden, Dresden, Germany
Abstract :
In this contribution we present a CNN video processing system that takes full advantage of a digital CNN Universal Machine (CNN-UM) using the recently-introduced NERO architecture. The proposed system is fully configurable in terms of network size and data precision. Using a Xilinx Zynq-7000 SoC we implemented a network of 512 × 900 cells at 8 bit, enabling us to process video streams with more-than-VGA resolution in real-time.
Keywords :
field programmable gate arrays; system-on-chip; video signal processing; CNN-UM; NERO architecture; Xilinx Zynq-7000 SoC; cellular nonlinear networks; digital CNN Universal Machine; high-speed real-time video processing platform; video stream processing; Arrays; Field programmable gate arrays; Program processors; Random access memory; Real-time systems; Streaming media;
Conference_Titel :
Cellular Nanoscale Networks and their Applications (CNNA), 2014 14th International Workshop on
Conference_Location :
Notre Dame, IN
DOI :
10.1109/CNNA.2014.6888618
