- Additional Service for FPGA and Quick Evaluation Available with "DE10-Nano" Board -

LeapMind Inc., a leading deep learning solution provider for enterprises, announced on July 17 that "DeLTA-Lite," a building solution for embedded deep learning models without programming, to officially support "Cyclone (R) V SoC" delivered by Intel Corporation of the United States. "DeLTA-Lite" started to provide an additional, FPGA-oriented service from July 2, 2018.

The said FPGA is loaded in "DeLTA-Kit," a hardware kit for easily evaluating deep learning released on June 27, 2018.

(Image: https://kyodonewsprwire.jp/img/201807125959-O1-5m81xYZ9)

At the release of "DeLTA-Lite" in April 2018, the scope of its support was for CPUs and FPGAs, but now expanded its scope to "Cyclone (R) V SoC" as well to reach out to more clients. LeapMind is planning to diversify hardware support for "DeLTA-Lite" in the future.

The kit released in June will enhance usability of "DeLTA-Lite" by enabling an additional feature where it generates downloadable binary and configuration files after training deep learning models. Those outputs can be executed on FPGAs.

About "DeLTA-Lite"

"DeLTA-Lite" is a cutting-edge solution to make it possible to build and deploy embedded deep learning models for practical use.  Until now, introducing embedded deep learning has required highly specialized knowledge and skills both for model designing and hardware implementation. However, by using "DeLTA-Lite," only a few steps are required to build embedded deep learning models. It also significantly reduces time and cost to deploy those models onto a small edge device, allowing the implementation of a detection function into a small machine or a robot, which otherwise could not be realized without them.

DeLTA-Lite official website: https://delta.leapmind.io/lite/en/

About LeapMind Technology

LeapMind provides one-stop solutions from model building and model compression to model implementation onto hardware so that deep learning technology can be enjoyed within a small computing environment where access to electricity is limited.

(1) Unique Deep Learning Algorithm

LeapMind conducts research on its own innovative algorithms that can reduce the computational complexity of deep learning to use within a small computing environment such as FPGAs.

(2) Optimal Hardware Architecture for Deep Learning

LeapMind also conducts research on original chip architectures that can efficiently implement deep neural networks on a circuit such as FPGAs with low power and limited memory.

LeapMind is making continuous efforts to make deep learning “small and compact” and accessible across a broad spectrum of applications, evolving the Internet of Things into the “Deep Learning of Things (DoT).”

About LeapMind

Head Office: Shibuya-ku, Tokyo

Representative: Soichi Matsuda, CEO

Established: December 2012

URL: https://leapmind.io/en/