New Computer Vision SDK Beta R2 Brings Enhanced Deep Learning Capabilities, Supports More OSs, & Improves Performance

Just Released! Intel® Computer Vision SDK Beta R2

For computer vision, deep learning, video solutions (such as security cameras), or office automation, top systems may incorporate multi-channel streaming, real-time software-based analytics, and more. On the heels of its newest release, you can get your geek on to integrate intelligence into your apps with the updated Intel® Computer Vision SDK (Intel® CV SDK) Beta R2. It delivers improved runtime performance, code algorithm efficiencies, and enhancements to the Deep Learning Deployment Tool. And we can’t forget – new support for more OSs: Windows 10*, Ubuntu 16.04*, OpenCV 3.3*, Yocto MR3*. Additionally, a new easy-to-install and use OpenCL driver is included.

Dive into the Fun Technical Stuff

See below the details of what's new in this release.

Improved Deep Learning Tools

Model Optimizer

• New support for TensorFlow*, a deep learning framework (partial features apply to GoogleNet V1, V3).
• Horizontal fusion support for Caffé.

Inference Engine

• Custom Layers support on GEN.
• Provides an application to verify the actual inference on the target hardware to validate accuracy (top1/top5) on image classification problems and to measure performance (in FPS) for selected batch sizes.

New topologies are now supported

• SqueezeNet, GoogleNet (Caffee* V2,V3), Resnet (50, 101, 152), SSD, YOLO1

Improved Runtime Performance

• Improved performance of our Inference Engine on CPU and our GPU/GEN Graphics processor.
• Intel® CV SDK OpenVX Runtime internal optimizations make computer vision building blocks more efficient – reducing development time-to-market.
• Improved scheduling with Intel® Threading Building Blocks closed the gap for heterogenous use cases and give you the ability to control where your code is running easily (CPU, GPU and so on) without writing targeted code for each hardware platform.
• Intel® Math Kernel Library for Deep Neural Networks (Intel® MKL-DNN) integration and CNN extension fusion improves the deep learning workloads performance.

New Features in the Vision Algorithm Designer (VAD)

• The new version supports OpenVX legacy code and provides the ability to upload legacy OpenVX code and generate Graphml for backwards compatibility.
• Application tracing and performance analysis capabilities help you as a developer improve code performance.
• Deep learning application support helps you to better understand optimized deep learning OpenVX Graphs.
• Model Optimizer generated code in VAD provides the ability to debug the application.
• OpenCL™ Custom kernel code generation is now supported.

Additional OS Support, OpenCL driver Now Available 

The new release includes support for Windows* 10, Ubuntu* 16.04, OpenCV* 3.3, and Yocto* MR3. Additionally, a new easy-to-install and use OpenCL driver is included.

The Computer Vision SDK supports these Intel® platforms:

• 6th generation Intel® Core™ processors
• Next-Generation Intel® Atom  processors

Download the Intel Computer Vision SDK now.

About the Intel Computer Vision SDK

The Intel Computer Vision SDK is a comprehensive toolkit for developing and deploying computer vision solutions on Intel® platforms, including digital surveillance cameras, robotics, autonomous vehicles, office automation, mixed-reality headsets, and more. Based on OpenVX*, this SDK offers many useful extensions and supports heterogeneous execution across CPU and SoC accelerators using an advanced graph compiler, optimized and developer-created kernels, and design and analysis tools. It also includes deep learning tools that unleash inference performance on deep learning deployment.

For more such intel IoT resources and tools from Intel, please visit the Intel® Developer Zone

Source:https://software.intel.com/en-us/blogs/2017/07/14/new-computer-vision-sdk-release-delivers-faster-performance-code-algorithm