fbpx
Au-Zone Logo EVA

Unleash the full potential of Deep Learning on your platform

Calgary, May 1, 2017 /PRNewswire/ — Today, Au-Zone Technologies Inc. announced the introduction of DeepView, a Machine Learning Toolkit and Run-Time Inference Engine for embedded processors.

The DeepView Toolkit provides computer vision engineers with a seamless and intuitive, graphical development environment to design, train and deploy Convolutional Neural Network solutions on a wide range of embedded processors. DeepView is integrated with TensorFlow, Google’s training framework, and fully compatible with cuDNN, NVIDIA’s training acceleration library, for desktop machines with compatible GPUs. By providing a full-featured development environment with best-in-class 3rd-party integrations, DeepView allows developers to design and test custom CNN networks on proprietary hardware platforms within hours.

The range of applications where Deep Learning techniques are being applied and providing commercial value has been growing exponentially since its rediscovery in the 2012 ImageNet competition. As research continues in the academic world and cloud based commercial applications flourish, the desire to employ the same methods and techniques on general purpose embedded processors has also grown. As with any new technology, the challenge to implementation is often finding experienced resources, access to relevant tools and adequate time for development – DeepView addresses these three issues directly.

“With the DeepView ML Toolkit and hardware accelerated runtime inference engine that we’re introducing today, developers with limited time or computer vision experience will be able to quickly and efficiently design, train and deploy advanced Deep Learning solutions on a wide range of embedded processors,” said Brad Scott, President of Au-Zone. “We see a great potential to help development teams, OEMs and technology vendors to deliver a wide range of unique, market specific object detection and image classification solutions using Deep Learning on general purpose SoC processors.”

The DeepView Toolkit serves a wide range of markets and the scope of applications is only limited by your imagination and relevant datasets for training. Some specific examples

  • ADAS and eCockpit for consumer vehicles
  • Heavy equipment and agriculture platforms
  • Healthcare and patient monitoring
  • Business and retail analytics
  • Mobile and Wearable devices
  • Security and Surveillance
  • Industrial machine vision
  • Consumer electronics

“The DeepView ML Toolkit allowed NXP and Au-Zone to quickly and efficiently develop and deploy a CNN-based Traffic Sign Recognition solution on NXP’s latest i.MX8 application processor,” said Rafal Malewski, Head of the Graphics Technology Engineering Center at NXP. “By using their ML Toolkit, and runtime inference engine which is optimized for i.MX GPGPU, we were able to assess real world performance on the development platform within a few days after starting the project”

DeepView ML Toolkit

  • Feature-rich dataset curation workspace with support for still image and video imports
  • Graph-based network design with several ready-made network templates and 15+ fully configurable primitives to simplify custom network designs
  • Fully integrated training with TensorFlow and support for cuDNN on machines with NVIDIA GPUs
  • User-configurable image augmentation feature quickly enhances dataset size and quality
  • Full featured network validation suite with real-time tools for on target runtime profiling and optimization
  • Support for common desktop environments including: Windows, Linux and Mac.

DeepViewRT

  • Production-ready inference engine with hardware acceleration & architecture-specific optimizations for ARM v7/v8 (NEON), x86, and other GPUs and FPGAs with OpenCL 1.2 support
  • Heterogeneous processor support for CPU, GPU, and proprietary Vision Processors (VPUs)
  • Target-optimized versions available for NXP i.MX
  • Model compression, encryption and authentication for over-the-air updates to field-deployed devices
  • Support for Linux and Android, with future support for additional RTOSs

About Au-Zone Technologies Inc.

Au-Zone Technologies Inc. is a leading provider of development tools, engineering design services, and enabling IP used for the design of intelligent embedded vision products and solutions. By utilizing our embedded development tools (eCV SDK and DeepView Machine Learning Toolkit) we enable engineering teams to quickly develop and deploy hardware accelerated computer vision algorithms and novel Convolutional Neural Networks on heterogeneous compute devices. Through our engineering consulting engagements and turnkey product development services, we help our clients lower development costs, mitigate program risk and shorten time to revenue when designing new vision enabled products. As an ecosystem partner, these development tools, design services and related IP help technology vendors to broaden their market opportunities and better serve their customers. Visit us at www.embeddedcv.com and follow us on YouTube, LinkedIn, and Twitter.

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 North California Blvd.,
Suite 360
Walnut Creek, CA 94596 USA

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