3DQ : Depth video compression and streaming library

3D VIDEO STREAMING AND COMPRESSION

Aivero has developed a high performance, world leading 3d depth compression and streaming software that gives our customers the power to recreate the real world in the cloud in real time by using off the shelf, low cost depth and 3d cameras.

Aivero has developed a high performance, world leading 3d depth compression and streaming software that gives our customers the power to recreate the real world in the cloud in real time by using off the shelf, low cost depth and 3d cameras.

Slider
20/1 compression ratio

HD Video, full frame rate

Little CPU load

Real Time 3D Video Streaming

Looking to scale 3D video computer vision or volumetric video applications? Once you start dealing with 3D volumetric video you will quickly realize that large data rates might represent an issue.

Another problem is the limited amount of processing you have for running deep learning training of data on local hardware. Training deep neural networks requires significant processing and the data needs to be transported to cloud computers where clusters of CPUs units, TPUs (Tensor Processing units) and GPUs (Graphical Processing Units) can execute the training. To transport the data and store it in the cloud, you will need compression and streaming.

Then there is the problem of accessing real time data from Depth video cameras over local networks, mobile networks and internet connections. The applications here could range from 3D video surveillance, Holographic telepresence via VR, to control of Drones, Robots and Autonomous vehicles. Today, most depth cameras are connected locally and all real time processing is done locally. As the algorithms and AI technologies grow more advanced, they also require more processing power. At the same time many of these applications require cost optimization, small form factor hardware as well as fanless operation. This means that the data needs to be transported to the cloud not only with adequate bandwidth, but also with extremely low latency.

Read more
Hardware acceleration on Intel, NVIDIA, ARM

Simple C++ API, bindings for Python and other languages

Integrations with GStreamer and openCV

Compress Store Stream

In Aivero we have addressed problems related to depth video data and created an algorithm that is able to compress depth video to very low data rates without significant loss. Aivero’s 3DQ technology features efficient compression of depth video in real time using existing hardware acceleration found in most computers, mobile phones and embedded CPUs. This enables streaming RGB and Depth video (RGB-D) at high frame rate and resolution on a wide range of desktop, embedded and mobile devices. 3DQ offers hardware accelerated depth video compression for any 3D camera. Efficiently compress any depth-map or single PoV point-cloud Store it locally and access it using a simple API or stream it with low latency across Ethernet, wifi or mobile networks.

Compress Store Stream

In Aivero we have addressed problems related to depth video data and created an algorithm that is able to compress depth video to very low data rates without significant loss. Aivero’s 3DQ technology features efficient compression of depth video in real time using existing hardware acceleration found in most computers, mobile phones and embedded CPUs. This enables streaming RGB and Depth video (RGB-D) at high frame rate and resolution on a wide range of desktop, embedded and mobile devices. 3DQ offers hardware accelerated depth video compression for any 3D camera. Efficiently compress any depth-map or single PoV point-cloud Store it locally and access it using a simple API or stream it with low latency across Ethernet, wifi or mobile networks.

Read more

Specifications

  • Linux
  • Windows
  • Android

Up to 1080p at 30fps

  • Typical compression ratio of 20/1
  • ~ 30% CPU load for compression and UDP transport of:
    • 1080p 30fps depth video
    • 1080p 30fps RGB video (simultaneously)
  • C++14
  • Lossy compression with variable compression ratio – Currently up to 1080p at 30fps (4K soon)
  • Simple interface using std containers
  • Bindings coming soon:
        – Rust
        – Python
  • CUDA preprocessing
  • Docker
  • Gstreamer
  • WebRTC