Sorry, you need to enable JavaScript to visit this website.

Open Visual Cloud – Updates and What’s New

BY Xintian Wu ON Aug 05, 2019

Open Visual Cloud – Updates and What’s New

Tao Tao Q, Jindu Hou and Jun Tian contributed to this article. 

On April 8, 2019, Intel launched the Open Visual Cloud, an open source project dedicated to providing optimized ingredients and sample reference pipelines for visual cloud services (media delivery/processing, media analytics, immersive media, cloud gaming and cloud graphics). Since then, there have been some major additions and improvements we wanted to share with you.

SVT Encoder Core

The Open Visual Cloud is made up of various open source components, assembled into pipelines to preform one of the five Visual Cloud services. You can read all about these services on this very website by clicking the tab “Pipelines and Services”.
 
Scalable Video Technology (SVT) is a major component of the Open Visual Cloud and is included as the software encoder in every Open Visual Cloud pipeline. SVT is an Intel optimized architecture for open source video encoder cores (not a full product), which currently comes in three flavors: SVT-HEVC, SVT-AV1, and SVT-VP9. To put it more plainly, the SVT video encoder core is able to take a source video format and encode it into HEVC, AV1 or VP9 for streaming over the internet. How it does this and why you should care can be researched at the SVT page on 01.org.
 
However, this blog post is about what’s new, so let’s get on with it!

SVT-AV1  0.6.0 

The SVT engineering team has been hard at work banging away on their keyboards late into the night, and the big news is we have a new, numbered SVT-AV1 release, SVT-AV1 0.6.0!

SVT-AV1 0.6.0 was released on June 28, 2019 and contains the following updates:

  • Initial decoder implementation
  • Static library support
  • Alt-ref pictures - temporal filtering
  • Adaptive Transform Block for INTRA
  • Adaptive QP scaling
  • Decoder - Support for Tiles and 10bit
  • API - add option to calculate / report PSNR values
  • Support for segmentation
  • SIMD Optimizations
  • Downsampling 2x2 filtering
  • Handle incomplete SBs
  • MACROS / trailing / tabs-spaces cleanup
 
More information can be found at the SVT-AV1 GitHub* page, here. Also, visit our SVT-AV1 Trello board to see what is coming.  

SVT-HEVC  1.4.0

Not to be outdone by AV1 updates, the SVT team has released the latest version of SVT-HEVC! SVT-HEVC 1.4.0 was posted to our Github repository on August 9, 2019. SVT-HEVC is much more mature than SVT-AV1 and v1.4.0 contains the following updates:
  • Deprecated-tune interface (only legacy tune 1 remains valid)
  • Added VBV (Video Buffering Verifier) feature
  • Added Y4M file support
  • Added support for YUV422/YUV444 with 8bit/10bit color formats
  • Added GStreamer plugin
  • Added motion-constrained tile sets (MCTS) feature
  • Added support for motion vectors to be constrained to picture boundaries
  • Added HDR (High dynamic Range) feature
  • Improved thread shutdown handling
  • Improved build process and support for VS 2019
  • Validated on CentOS
  • Added Style Guide
  • Expanded test documentation
  • Numerous bug fixes including
    • Eliminated corruption of interlaced input
    • Rate distortion computation
    • Eliminating Compiler warnings
    • Bit estimation overflows
    • Fix differences in bitstream between intrinsic/c code
    • Improved variable bit rate (rc=1) encoding stability

You can directly download SVT-HEVC v1.4.0 at the SVT-HEVC v1.4.0 Github repository at this link.

Open Visual Cloud – 2 new services reference pipelines and multiple sample updates!

The Open Visual Cloud team has created two new services pipeline samples. This new Open Visual Cloud release also includes new tutorials to help developers get started. 

Video Conferencing Service and Sample 

The video conferencing sample (not a finished product) implements a web meeting demo based on Open WebRTC Toolkit (OWT) media server and client SDK, which fully demonstrates OWT media processing features in both mix and forward modes. It also provides basic conferencing actions like screen sharing, instant messaging, and meeting control in web UI.
 
The OWT media service is delivered as separately as a Docker* based image, ready for application extension or deployment. Learn more at the Dockerfile GitHub repo.
 
You can also learn more about the Video Conferencing Sample at the Video Conferencing Sample GitHub repo or get started coding here: Tutorial - Build and Run the Video Conference Sample on public cloud or local machine.

Smart City Traffic Management Sample

The smart city traffic management reference pipeline shows how the integration of the various media building blocks, including SVT, with analytics powered by the OpenVINO™ Toolkit for smart city use cases including street-corner traffic control (city planning) can be used as a starting point. This sample (not a finished product) can be referenced by developers to ease application development challenges. It enables real time analytics of live video feeds from IP cameras.
Just like the Video Conferencing Sample, we have a couple of links for you.

Content Delivery Network Transcode Sample - New Documentation

In an effort to continuously upgrade our developer focused documentation, we have created a series of tutorials around the CDN Transcode Sample. You can find these under the Wiki tab on the CDN Transcode Sample GitHub repo and also at the direct links below.
 

Intelligent Ad-Insertion Sample - Optimization for 2nd Generation Intel® Xeon® Scalable Processors

The Ad-Insertion sample added support for 2nd Generation Intel® Xeon® Scalable Processors (formerly Cascade Lake). The sample utilizes the OpenVINO toolkit capabilities to speed up the INT8 model execution on the Cascade Lake platform, improving inference performance. The sample additionally provides Kubernetes scripts if you wish to deploy in a Kubernetes cluster.
 
Whew, that is a lot of updates! 
 
The team is excited to see developers try out or sample pipelines and we would love to see what you are building. If you want to share, please post your projects to the Open Visual Cloud topic in DevMesh, or share a link on Twitter and include “Open Visual Cloud” in your tweet so we can find it. 
 
OpenVINO and Xeon are trademarks of Intel Corporation or its subsidiaries in the U.S. and/or other countries.