Processor IP

The videantis processors are the most efficient deep learning, embedded vision, and video coding processing solutions on the market. Its subsystems run all visual computing tasks on a single software-programmable and scalable processor architecture.

Processors that do more

videantis’ efficient processing subsystems run deep learning, computer vision, video coding, and imaging on the same compute platform. This reduces power consumption, bandwidth, and simplifies software development and integration.

To efficiently run deep convolutional nets in real-time requires extreme performance levels and careful optimization, which videantis addresses with its v-MP6000UDX processor IP and optimization tools.

 

Unified architecture

The unified architecture runs all visual processing tasks.
This saves power and silicon area, reduces time-to-market, and extends product life.

Easy hardware integration

videantis provides full subsystems that include a memory hierarchy, multiple processors, a bus fabric or network-on-chip, and various system interfaces. The processing subsystems are ready for integration into SoC designs with minimal design work and proven power, performance, and area results.

Single integrated software tool flow

Whether it’s deep learning, computer vision, imaging, codecs, or generic algorithm acceleration using OpenCL, all software components are developed using a single programming toolset. This reduces the development effort and bugs, and maximizes performance.

Minimize data bandwidth

Instead of moving data around between accelerators and different processor subsystems, everything is kept local in the v-MP6000UDX subsystem. This lowers bandwidth and ensures that compute engines will not be starved for data any more.

Deep learning

Deep learning and AI are transforming industries.
Whether it’s at the edge or in the cloud, videantis has the solution to bring deep learning into the consumer’s hands.

High performance

Whether it’s at the edge or in the cloud, videantis' processors provide the required performance.

Scalable architecture

1 to 256+ cores address ultra-low cost to high-performance applications.

Ultra-high MAC throughput

Each core computes 64MACs per cycle, resulting in a total of 16,384MACs per cycle (for 256 cores) or even more.

Enhanced memory architecture

A distributed and hierarchical memory architecture combined with a high-performance bus fabric results in ultra-high-bandwidth and low power.

Runs complete neural nets

All layers of CNNs are computed on the same programmable processor, not just the convolutional layers.

Software programmable

Smaller and lower power than hardwired solutions, yet fully software programmable; runs any neural network efficiently and is future proof.

Run all networks with ease

The v-CNNDesigner tool automatically translates neural networks into parallelized and optimized multicore code for the videantis processing platform.

Runs and optimizes every CNN in minutes.

Supports TensorFlow, PyTorch, Caffe, ONNX, and more.

Supports all key networks
including ResNet, YOLO, MobileNet, VGG, and Inception.

Computer vision

Seamless acceleration for APIs such as OpenCV, OpenVX, or proprietary algorithms.

Сomputer vision acceleration libraries

The videantis comprehensive computer vision acceleration libraries include seamless acceleration for APIs such as OpenCV and OpenVX.

Proprietary algorithms

Proprietary algorithms can be optimized using videantis' state-of-the-art toolchain with fast multicore simulator including host processor co-simulation support.

Image processing functions

videantis offers a large set of image processing functions for offloading the host CPU.

Available functions include

  1. Object detection
  2. Optical flow
  3. SLAM/structure from motion
  4. Feature detection and description
  5. Image transformations
  6. and many more

Video coding

videantis offers the most flexible, small area and low-power video coding solution on the market.

Looking to support multiple channels of high-resolution video encode and decode or upgrade to the latest standards?

Need low-delay, H.264 or JPEG, 10-/12-/14-bit codecs that run well over 100Mbps for automotive applications?

videantis’ extensive library of 15+ video and still image codecs addresses a wide variety of video coding applications.

Development tools

Easy to use and complete software development kit to get products to market fast.

Available products include

  1. Eclipse-based IDE
  2. v-CNNDesigner maps and optimizes deep learning workloads
  3. videantis-C C compiler with vector support
  4. High-level code generator
  5. High-speed multicore simulator
  6. Profiler and debugger
  7. Production sign-off tool suite for automotive development and ISO 26262

Demo videos

Contact us

Street address

videantis GmbH
Rotermundstraße 11
30165 Hannover, Germany

Customer support

videantis’ industry-leading products are backed with top-rated service and support. Customers can log in to the videantis support center.

General info

Having a generic question about videantis or its products? The videantis team is here to help.

Public relations

Having a question for the videantis marketing communications team?