Qualcomm Technologies is teaming up with SenseTime to implant ‘deep learning’ abilities into IoT machines.
Qualcomm Technologies, a subsidiary of Qualcomm, has teamed up with with Hong Kong-based artificial intelligence (AI) specialist SenseTime, in order to integrate the latter’s machine learning models and algorithms with its own Snapdragon system-on-a-chip (SoC) semiconductors.
By joining forces, the companies hope to make new advances in on-device AI, in areas such as camera-based image processing and so-called innovative vision. In other words, this partnership is about the application of context-centric image processing on devices themselves, in order to determine subject matter and identify various scenarios out on the edge.
The on-device AI bandwagon
The move mirrors the on-device AI strategy recently announced by Chinese telecoms and cloud giant Huawei, detailed here on Forbes. By implementing AI on the device itself, rather than on a back-end server, these companies believe they can deliver significant advantages over cloud-only implementations.
The idea rests on their ability to enable edge relatively pared-down edge devices to reliably execute AI processes, with or without a network connection.
Dr Li Xu, co-founder and chief executive officer of SenseTime claims that the benefits of this include improved real-time performance, privacy protection and enhanced reliability.
“Qualcomm has been conducting fundamental research in AI over a decade,” said Keith Kressin, senior vice president of product management at Qualcomm Technologies.
“In fact, many devices shipping today using our Snapdragon mobile platforms already utilize on-device AI. We look forward to the results of our collaboration with SenseTime to further accelerate new and exciting capabilities of on-device AI for millions of customers using mobile devices.”
On-device AI use cases
Currently, Qualcomm is focused on optimizing the Snapdragon mobile platform to accelerate myriad AI use cases in the areas of computer vision and natural language processing (NLP) in smartphones, IoT devices and in the automotive sector.
The company says it is also researching broader applications in areas including wireless connectivity, power management and photography.
SenseTime, for its part, is playing a role in deep learning algorithm development and has built a proprietary deep learning platform called Parrots. The company’s technology in deep learning makes it possible to develop a variety of algorithms with quick turn-around. SenseTime has also made breakthroughs in algorithm model miniaturization.
In July this year, the company completed a $410 million funding round, claiming it was the largest private financing round ever closed by an AI start-up.
The combination of the two firm’s smarts has the potential to improve the speed and efficiency of combining algorithms and chipsets to make IoT AI technology more pervasive.
Coming soon: Our IoT Build events, taking place in London on 14 & 15 November 2017 and San Francisco on 27 & 28 March 2018 are a great opportunity for attendees to explore the platforms, architectures, applications and connectivity that comprise the IoT ecosystem.