fbpx

Morpho Announces Commercialization of “SoftNeuro”, One of the World’s Fastest Deep Learning Inference Engines, Accelerating the Practical Use of Deep Learning

2017-symlogo-h-4c

Tokyo, Japan –December 5, 2017– Morpho, Inc. (hereinafter, “Morpho”), a global leader in image processing solutions, announced today the commercialization of “SoftNeuro™”, one of the world’s fastest1 deep learning2 inference engines. Morpho not only licenses the engine alone, but also brings about a large increase in processing speeds of their existing engines such as “Morpho Deep Recognizer™”, which is Morpho’s image recognition engine, by embedding “SoftNeuro™” into them.

Recently, the development of the deep learning technology has been promoted for artificial intelligence. It has been successfully incorporated into services and products in a variety of fields and its practical use is progressing. However, opportunities to execute inference processing are increasing in the edge environment, and it is evident that there is an issue with the constraints of operating environments and inference speed. Therefore, Morpho developed “SoftNeuro™”, a fast inference engine that operates in many environments by utilizing learning results that have been obtained through a variety of deep learning frameworks.

“SoftNeuro™” is a general-purpose inference engine and is available for voice recognition and text analysis as well as image recognition. Therefore, not limited to image recognition, developer licensing is planned for a variety of inference engines that use deep learning.

“SoftNeuro™” is scheduled to be displayed and demonstrated at Morpho’s booth at the INTERNATIONAL TECHNICAL EXHIBITION ON IMAGE TECHNOLOGY AND EQUIPMENT 2017 to be held on December 6 to 8, 2017.

Features of “SoftNeuro™”, a deep learning inference engine

  1. One of the fastest in the world (According to Morpho’s research as of December 5, 2017)1
    In some applications of inference processing with deep learning, in some cases there is the issue of lengthy inference processing time.
    “SoftNeuro™" is faster than or equally fast as the mainstream inference engines – when run on CPUs – while providing a set of key features as described later. This high-speed performance has been achieved through a variety of optimizations (neural network, memory usage and others) for each platform.
    Please refer to note 1 for details.
  2. Supports multiple frameworks
    There are many frameworks that perform deep learning, including Caffe, Keras and TensorFlow™, which are open source software.
    “SoftNeuro™” achieves fast processing by utilizing the learning results of these major frameworks (Figure 2). It is possible to achieve fast inference processing (the first characteristic) and multi-platform compatibility (the third characteristic) without wasting the learning assets that users have built up so far. Further, the compatibility of “SoftNeuro™” with frameworks and layers will be expanded sequentially.
  3. Multi-platform compatibility
    Inference processing using deep learning is widely being applied to a variety of places, including smartphones, vehicles and FA equipment, not limited to cloud servers. In these operating environments, different platforms are used, including CPUs and OSs, therefore porting and optimization are necessary.
    “SoftNeuro™” is scheduled to be applied to a variety of platforms, and appropriate optimization (CPU speed-up instructions, use of GPU and DSP and others) for each platform will be performed.
    Multi-platform compatibility enables a flexible response to changes in operating platforms as well as expanded learning results for a wide range of operating platforms.
  4. Compatible with secure file formats
    Inference using deep learning is widely being applied to a variety of places, including smartphones, vehicles and FA equipment, not limited to cloud servers. To achieve their operation, a network is copied to many places after learning. This increases the risk of leaking learning know-how or results (original network structure, weight parameters and others).
    “SoftNeuro™” is capable of encrypting the trained networks, minimizing the risk of leaking the machine learning know-how and the results of learning.

Note 1: One of the fastest in the world (According to Morpho’s research in December 2017)

We compared the inference speed of “SoftNeuro™” with that of widely-used inference engines that we could obtain, on ARM and Intel CPU architectures (“SoftNeuro™” does not utilize GPU). Table 1 summarizes the evaluation conditions. Figures 1-1 and 1-2 present the results of the evaluation for these two architectures.


Table 1. Conditions for Processing Speed Comparison

Figure 1-1. Comparison of inference speeds (CPU: ARM. Measured on Qualcomm Snapdragon 835. Bars were not plotted for models that could not be implemented using a given inference engine, on this CPU.)

Figure 1-2. Comparison of inference speeds (CPU: Intel. Measured on Intel® Core™ i7-6700K CPU @ 4.00 GHz. Bars were not plotted for models that could not be implemented using a given inference engine, on this CPU.)

Note 2. About Deep Learning

Deep learning is a learning method in which artificial intelligence extracts characteristics from learning data, unlike conventional machine learning, which performs recognition based on characteristics defined by a human. Deep learning has mainly been applied to image recognition, voice recognition and language processing. It is popular technology that is expected to be utilized in a variety of fields, including marketing, security, automatic translation and automated driving.

Figure 2. Integrating “SoftNeuro™” into machine learning systems

About Morpho, Inc.

Established in 2004, Morpho is a research and development-led company in image processing technology. It has globally expanded its advanced image processing technology as embedded software, for domestic and overseas customers centered on the smartphone market, broadcasting stations and content providers. It has also provided image recognition technology utilizing Artificial Intelligence (AI), collecting image information captured by cameras into devices and clouds and analyzing it, for fields such as automotive devices, factory automation, and medical care. Morpho will provide broad support, making a wide range of innovations happen with its imaging technology and Deep Learning technology. For more information, visit http://www.morphoinc.com/en/ or contact [email protected].

Here you’ll find a wealth of practical technical insights and expert advice to help you bring AI and visual intelligence into your products without flying blind.

Contact

Address

1646 N. California Blvd.,
Suite 360
Walnut Creek, CA 94596 USA

Phone
Phone: +1 (925) 954-1411
Scroll to Top