CEVA DSP brings AI, deep learning compute performance to low-power embedded devices

SANTA CLARA, CA. CEVA, Inc. today introduced the CEVA-XM6-based vision platform, a digital signal processor-based (DSP-based) offering that brings deep learning and artificial intelligence (AI) capabilities to low-power embedded systems. The CEVA-XM6 imaging and vision DSP is the centerpiece of an integrated hardware and software silicon IP platform that allows developers to harness neural networks and machine vision for smartphones, autonomous vehicles, surveillance, robots, drones, and other camera-enabled smart devices.

The CEVA-XM6 delivers up to 8x higher performance for neural network workloads and up to 3x performance improvement than its predecessor, the CEVA-XM4, introducing new vector and scalar processing units and enhancements to instruction set, memory bandwidth, and direct memory access (DMA) through a parallel scatter-gather memory load mechanism that improves the performance of vision algorithms such as SLAM and depth mapping. A 3D data processing scheme for accelerated convolutional neural network (CNN) performance and 16-bit CEVA deep neural network (CDNN) accelerator round out the CEVA-XM6’s key features, while an optional 32-way SIMD vector floating-point unit (FPU) can also be integrated.

The integrated CEVA-XM6-based vision platform integrates software and hardware IPs that provide power consumption advantages for deploying machine vision and deep learning in embedded systems. The platform includes function-specific accelerators for CNN and image de-warp, CEVA’s CDNN2 neural network software framework, OpenCV, OpenCL and OpenVX APIs, CEVA-CV computer vision library, and a set of optimized algorithms for programming flexibility.

“As computer vision and deep learning technologies become mainstream, there is a need to bridge the gap between the deep neural networks that are being generated by power-consuming GPU engines and the ability to deploy these in power- and performance-constrained embedded applications,” said Ilan Yona, Vice President and General Manager of the Vision Business Unit at CEVA. “Our new vision platform excels in this regard, providing developers with the most comprehensive set of technologies to rapidly address these embedded use-cases.”

The CEVA-XM6 DSP and vision platform components will be available for licensing to lead customers in the fourth quarter of 2016 and for general licensing in the first quarter of 2017. For more information on the CEVA-XM6 DSP and vision platform, visit http://launch.ceva-dsp.com/CEVA-XM6.