ACCESS CO., LTD. announced today that , based in Tokyo, has entered into a collaboration agreement with ACCESS as an artificial intelligence (AI) technology partner for NetFront™ EdgeAI, a comprehensive solution that integrates edge-computing technology with AI technology to achieve high-speed, real-time AI processing on edge devices.
The number of IoT devices is increasing globally. It is predicted that approximately 30 billion IoT devices*1 will be connected by 2020, ranging from high-spec and high-end devices such as connected cars to low-end devices with limited power consumption and footprints such as surveillance monitors, small cameras and drones. The market expects AI technology to be combined with the ever-increasing number and types of IoT devices to drive take-up and effectiveness. Demand is expected to be particularly high for the provision of tools to implement AI technology to achieve real-time image recognition on IoT edge devices and manage them in an efficient, secure and safe manner.
In light of this, ACCESS and LeapMind have entered into a collaboration agreement to provide a solution that combines ACCESS’ advanced technology for managing edge devices in the cloud remotely with LeapMind’s advanced model compression technology. LeapMind’s advanced model compression technology equips small computing environments, including low power consumption SoC FPGAs, with deep-learning inference algorithms.
Specifically, the combination of ACCESS’ NetFront™ EdgeAI with LeapMind’s DeLTA-Lite will be provided as a comprehensive solution for building deep-learning models, distributing them to each edge device and updating them, thereby helping companies efficiently deploy and operate IoT solutions. Targeting low-cost, low-power consumption edge devices, this solution makes it possible to implement recognition features on various edge devices such as small cameras that were not previously able to utilize deep learning.
Combining ACCESS’ NetFront™ EdgeAI with LeapMind’s DeLTA-Lite
Soichi Matsuda, CEO of LeapMind, said: “I am sure that this collaboration with ACCESS will accelerate the implementation of Deep Learning of Things (DoT), in which deep learning technology is applied to all kinds of objects. Going forward, LeapMind will strive to make complex and complicated deep learning “compact and simple” and embed it to the real world.
Michi Uematsu, CTO of ACCESS said: “This collaboration with LeapMind makes it possible to actively implement AI technology on embedded devices. ACCESS has been continuously transforming society by providing Internet technology to various embedded devices. By providing IoT devices powered by LeapMind’s AI technology for embedded devices and ACCESS’s edge-computing technology for embedded devices, we will keep improving everyone’s daily life.”
ACCESS and LeapMind will provide this integrated solution to several companies in advance as a pilot test and will offer it officially by the end of the year to factories, warehouses, logistics, security and other fields that are working to equip their numerous IoT devices with AI.
Examples of solutions that combine the two companies’ technologies
[Roles of the two companies]
Provide NetFront™ EdgeAI, a solution for dramatically reducing processes in the cloud and achieving fast, real-time processing that is not dependent on any network environment, by equipping edge devices with AI features.
Provide DeLTA-Lite, a solution for building embedded deep learning models that does not require programming
NetFront EdgeAI: Composed of NetFront™ Agent and ACCESS Connect™. NetFront Agent is one of the world’s smallest managed computing engines and was developed based on the proprietary engine of NetFrontTM Browser, a well-established product that has been implemented in more than 1.5 billion devices around the world. ACCESS Connect is a one-stop platform that provides backend systems required for developing and operating IoT services.
Learn more about NetFront Agent:
DeLTA-Lite: A web-based solution for building embedded deep learning models that does not require programming.
This makes it possible to run deep learning models on palm-sized devices by simply preparing training data and hardware for implementing it. With DeLTA-Lite, you do not need to write a program code, thereby shortening the time to complete the process from building a model to verifying it on hardware.
Learn more about DeLTA-Lite:
*1 Edited based on the source ”White Paper by Information and Communications in Japan Year 2017.”
Since 1984, ACCESS CO., LTD. (Tokyo Stock Exchange Mothers Index, 4813) has provided advanced IT solutions centered around mobile and network software technologies to telecom carriers, consumer electronics manufacturers, broadcasting and publishing companies, the automotive industry, and energy infrastructure providers around the world. The company develops mobile software solutions that have been installed on over 1.5 billion devices, and network software solutions that have been used by over 300 telecommunication equipment manufacturers. Utilizing its network virtualization technology skills and knowledge, the company is currently focusing on the development and commercialization of Internet of Things (IoT) and media solutions that combine embedded and cloud technology. Headquartered in Tokyo, Japan, the company operates subsidiaries and affiliates in Asia, Europe and the United States to support and expand its business globally. Learn more about ACCESS at www.access-company.com
About LeapMind Inc.
LeapMind conducts research and development of deep learning solutions for embedded devices and achieve Deep Learning of Things (DoT), in which deep learning is applied to all types of objects. By improving neural network models and conducting independent algorithm research for both software and hardware, the company has developed key technologies to run deep learning on edge devices even in a small computing environment with a limited power capacity, such as low power consumption FPGAs. LeapMind leverages this unique technology to provide one-stop solutions for building models, compressing/optimizing them, and implementing them on edge and embedded devices.