CEVA unveiled the second generation AI processor architecture NeuPro-S for edge deep neural network inference at AutoSens, Brussels, Belgium. In conjunction with the release of NeuPro-S, CEVA also launched the industry's first deep neural network compiler technology CDNN-Invite API, which optimizes neural network inference firmware to support heterogeneous collaborative processing of NeuPro-S kernel and custom neural network engine through a unified interface. . NeuPro-S, along with the CDNN-Invite API, is ideal for any vision-based device that requires edge artificial intelligence processing, including autonomous vehicles, smartphones, surveillance cameras, consumer cameras, and AR/VR headsets, robotics, and industrial applications.
Designed to optimize NeuPro-S, segmentation, detection and classification of neural networks in video and imagery in edge devices, significantly improving system perceptual performance. These include support for multi-level memory systems to reduce the high cost of transmission using external SDRAM, and support for multiple compression options and heterogeneous scalability to implement CEVA-XM6 vision DSP, NeuPro-S core and custom AI engine in a single unified architecture Various combinations. Compared to CEVA's first-generation AI processor, NeuPro-S has an average performance increase of 50%, memory bandwidth and power consumption reduced by 40% and 30%, respectively.
The NeuPro-S family includes the NPS1000, NPS2000, and NPS4000, which are pre-configured processors with 1000, 2000, and 4000 8-bit MACs per cycle. With the highest single-core CNN performance, the NPS4000 achieves 12.5 TOPS at 1.5 GHz and is fully scalable up to 100 TOPS.
For increasingly diverse specialized neural networks and processors, the CDNN-Invite API seamlessly integrates customer-designed neural network engines into CEVA's award-winning Deep Neural Network (CDNN) framework. CDNN will then fully optimize and enhance the network and layers to take advantage of the superior performance of the CEVA-XM6 vision DSP, NeuPro-S and custom neural network processors. The CDNN-Invite API has been adopted by major customers who work closely with CEVA engineers to deploy them in commercial products.
Ilan Yona, vice president and general manager of CEVA's visual business unit, commented: "CEVA continues to lead the development of deep edge neural network processors and optimization technologies for extended edge devices. The NeuPro-S architecture addresses the increasing data bandwidth and power consumption of these devices. The root cause of the challenge. Through the CDNN-Invite API, we have reduced the barriers to entry for the growing community of neural network innovators, allowing them to benefit from the extensive support and ease of use provided by our CDNN compiler, thereby further Extends the undisputed competitive advantage in the field of neural network compiler technology."
CEVA's extensive experience and overall strategy in integrated processing, acceleration, software and tools provides an open, efficient and programmable platform for deep learning. In addition to AI real-time processing, the fully programmable CEVA-XM6 vision DSP integrated in the NeuPro-S architecture can simultaneously process images, computer vision and general DSP workloads. This enables customers and algorithm developers to leverage CEVA's extensive library of image and visual software, including the CEVA-SLAM software development kit for 3D mapping, the CEVA-CV and CEVA-VX software libraries for computer vision development, and more recently The wide-angle image software suite includes de-warping, video stitching and image data sensor fusion technology.
NeuPro-S further strengthens CEVA's success in the automotive case by providing solutions that meet safety requirements, including quality assurance standards IATF 16949 and automotive standards including ISO 26262 and A-Spice.
CEVA has now released NeuPro-S and has licensed it to leading customers in automotive and consumer product camera applications. Currently CEVA has provided the CDNN-Invite API to leading customers and has a general license at the end of 2019.
CEVA introduces new edge device AI reasoning processor architecture to provide co-processing support for custom neural network engines
Feb
02
65