Ceva launches AI processor architecture and API for edge computer vision applications
Wireless connectivity and smart sensing technology provider Ceva has introduced its second-generation AI processor architecture, NeuPro-S, to support deep neural network inferencing at the network edge.
The company also launched the CDNN-Invite API, which provides deep neural network compiler technology to support heterogeneous co-processing of NeuPro-S cores with custom neural network engines in run-time firmware for unified neural network optimization. The company says the API represents an industry first, and is ideal for vision-based devices requiring edge AI processing, such as autonomous vehicles, smartphones, surveillance and consumer cameras, AR and VR headsets, and robotics.
NeuPro-S is designed to optimize segmentation, detection, and classification of objects from edge device input with neural network processing. Ceva says it delivers significant performance improvements with system-aware enhancements, such as support for multi-level memory systems, multiple weight compression options, and heterogenous scalability to enable various combinations of AI engines in a single unified architecture.
“CEVA continues to set the pace in proliferating deep neural networks processors and optimization technologies for edge devices,” comments Ceva Vision Business Unit Vice President and General Manager Ilan Yona. “The NeuPro-S architecture addresses the root causes of the growing challenges in data bandwidth and power consumption in these devices. With our CDNN-Invite API, we have further extended our indisputable competitive edge in neural network compiler technologies by reducing the entry barriers for the growing community of neural network innovators, allowing them to benefit from the breadth of support and ease of use our CDNN compiler technology offers.”
The NeuPro-S family includes preconfigured NPS1000, NPS2000 and NPS4000 processors with corresponding numbers of 8-bit MACs per cycle.
Ceva says its engineers are already working with lead customers to deploy the CDNN-Invite API in commercial products. NeuPro-S has already been licensed for automotive and consumer camera applications, and will reach general licensing availability by the end of the year.
The automotive market for biometric technology is forecast to surpass $300 million by 2024, though how much processing will be done at the edge, rather than in the cloud remains to be seen.