Author: Ames Osram

    Ames Osram (Swiss Stock Exchange stock code: AMS), the world‘s leading optical solution provider, and Quadric, the IP innovator of the end-side AI machine learning processor, announced a strategic cooperation. The two sides will jointly develop an integrated sensor module, combine the Mira series of visible and infrared CMOS sensors of the leading edge of Ames Osram with Quadric‘s new Chimera? GPNPU processor. These modules achieve ultra-low power consumption, and the integration of both sides will provide innovative intelligent sensing solutions for wearable devices, robots, industrial and security markets. The cooperative products made their debut at CES Las Vegas from January 5 to 8, 2023, and were demonstrated on the spot.

    Quadric Chimera General Neural Network Processor (GPNPU) can complete machine learning graphics processing and classic data parallel C++algorithms in a unified architecture, bringing advanced end-to-end AI functions for edge-based applications. Quadric‘s architecture provides efficient machine learning (ML) reasoning performance, but unlike other neural network accelerators that support a limited number of machine learning graphical operators, Quadric‘s solution also has general control and signal processing capabilities, thus combining the ideal attributes of NPU accelerators with digital signal processors (DSP). Quadric GPNPU can simultaneously run neural network graphics and C++code for signal preprocessing and post-processing, and its power consumption is significantly lower than other edge silicon solutions.

    Ames Osram Mira series CMOS image sensors focus on reducing sensor and system power consumption, while minimizing product size to maximize resolution efficiency. The product portfolio of Mira series sensors is constantly expanding and providing a variety of resolutions to meet the different application needs of high-performance global shutter image sensors. Many products have been released in this series, and more products with high or low resolution are being developed.

    Joost Seijnaeve, vice president of Ames Osram and general manager of CMOS image sensor, said: "This cooperation will integrate Ames Osram sensor and Quadric processing into a low-power module, opening up a broad new possibility for intelligent visual sensor deployment. Original equipment manufacturers (OEMs) 5W edge silicon chip or 15W board-level computing solution is no longer needed to perform intelligent sensing functions, including human posture, face recognition, object detection, object recognition, scene analysis, etc. Today, the cutting-edge intelligent vision function can be embedded in a single integrated sensor and processing module, and applied to many innovative products in the fields of automobile, consumption, industry and so on. "

    Mira image sensor is combined with Quadric GPNPU to integrate image acquisition and machine learning into a low-power module. According to the simulation operation of Quadric, this product can complete the work task that currently requires 5W or 10WGPU card with only a few hundred milliwatts. The resolution of Mira220 and future derivative products, combined with Quadric‘s calculation range of 1TOP to 16TOP scalable Chimera? The processor lineup will be able to provide multiple combinations for different applications.

    Veer Kheterpal, chief executive of Quadric, said: "Quadric is very excited to work with Ames Osram this time. This product enables device manufacturers to develop fully programmable intelligent sensing devices with ultra-low power consumption, thus opening a new pattern of machine learning in edge devices."