Reference ClientsReference Clients

PALTEK

Hacarus and PALTEK announce collaboration on artificial intelligence-based FPGA products

  • Joint Development
  • Manufacturing
This joined press release announces the collaboration between Hacarus Inc. (Head office: Kyoto City Nakagyo-ku, Representative Director: Kenshin Fujiwara), a developer of artificial intelligence solutions and PALTEK Co., Ltd. (Headquarters: Yokohama-shi Kohoku, Securities Code: 7587, President and Representative Director: Naohide Yabuki), a semiconductor trading company that supports consigned design and development of FPGA.

Through this collaboration, PALTEK will develop FPGA products and box computer products equipped with Hacarus’ AI engine “HACARUS-X-Edge”.

 

HACARUS-X-Edge is a lightweight artificial intelligence solution that does not use the current mainstream deep learning methodology.  Instead, it applies a sparse modeling technology which excels in extracting features from small quantities of data. PALTEK will provide artificial intelligence devices that meet the needs of higher speed, smaller size, and lower power consumption by installing HACARUS-X-Edge in FPGA and box computers.

 

As artificial intelligence is being introduced across all industries and fields, several problems of utilizing deep learning are recognized. For instance the need for large amounts of learning data or the inability to explain the results from deep learning algorithms. Other aspects are related to security issues when transmitting data including sensitive personal information to the cloud via the internet. These challenges are barriers to the introduction of artificial intelligence in the embedded equipment field and medical field.

 

Since the autumn of 2017, Hacarus provides the AI solution “HACARUS-X-Edge” as main product, which combines sparse modeling and machine learning to extract features from small amounts of data. Sparse modeling is good at extracting the causal relationships between data and it is fully transparent in a way, that humans can understand the judging process and results of the artificial intelligence. In addition, HACARUS-X-Edge is optimized to be lightweight, so it can operate artificial intelligence on a completely offline device.

 

While FA robot manufacturers are developing 3D robot vision systems, the need for AI in closed environments is increasing. Utilizing the AI solutions of HACARUS-X-Edge loaded FPGA packages developed from this collaboration makes it possible to use artificial intelligence within a closed environment without uploading data to the cloud via the internet.

 

The integration of HACARUS-X-Edge with FPGA will conclude in December 2018. We plan to provide a development environment for box computers with this embedded solution in February 2019.

 

Example use case

Below is an example of an AI technology based moving object detection, running on a camera-equipped edge terminal with an ARM (Cortex A 9) processor. By applying feature extraction from a small amount of data with sparse modeling, the solution can distinguish background and moving objects based on a few seconds of camera footage.

 

This enables the AI algorithm to learn and optimize itself to the environment on an edge terminal environment, without the need of learning beforehand. Specific solutions include automatic tracking of people, detection of intrusion into buildings or restricted areas, and human flow analysis. A similar solution can be realized using FPGA.

 

 

About PALTEK Corporation

Since its founding in 1982, PALTEK is offering Japanese semiconductor products to domestic and overseas electronics manufacturers as well as design consignment services as a customer development partner. We support the entire process,  from specification review to prototyping and mass production. Based on the corporate philosophy of symbiosis and diversity, PALTEK helps their customers to find and develop optimal solutions. For details about PALTEK, please visit the website https://www.paltek.co.jp.

Reference ClientsReference Clients

Subscribe to our newsletter

Click here to sign up