Vitis™ Video Analytics SDK (VVAS)

Turn Vision Into Powerful Insights

Overview

Build Your Own Intelligent Video Analytics Solution with VVAS

VVAS is designed to build and deploy intelligent video analytics applications and services on Xilinx target platforms or your own platforms. Being built on top of Xilinx Vitis™ and Vitis™ AI, it is based on the popular Gstreamer framework and provides commonly used plugins including video, pre-processing, deep learning, and post-processing. You are free to write C/C++ or Python code with VVAS to build end-to-end IVA pipelines.

  • Optimized for end-to-end throughput and latency for multi-stream and multi-stage applications
  • Offers extensive AI models support for popular object detection, segmentation, and classification tasks
  • Custom plugins for hardware acceleration and software processing
  • Migration between embedded and PCIe acceleration platforms with minimal effort  
vvas-stack

The Complete Software Stack

To Build and Deploy AI-Based Intelligent Video Analytics Applications

Performance icon
Adaptable icon
Productivity icon

Performance & Scalability

Deliver the best-in-class, end-to-end performance across edge, on-premise, and cloud

Adaptable & Flexible

Provides flexible analytics workflow and offers extensive AI models to meet your needs

Productivity & TCO

From rapid prototyping to full production with less development effort and deployment cost

Featured Video

Build Your Intelligent Video Analytics Applications With Ease

Best-in-Class Performance

End-to-end throughput KV260 image

Source: WP529, benchmarked with Uncanny Vision’s ANPR application with VVAS and Deepstream 5.0 

End-to-end throughput VCK5000 image

AI model: Yolov3(416x416) + 3 * Resnet18(224x224), benchmarked with VVAS and Deepstream 5.0

alveo-vs-nvidia-2

Partner Solutions

aupera-logo

 

Smart Retail/City/Building

  • Full video AI solutions at the edge
  • Smart city, smart building, and smart retail
  • High efficiency, low latency, and scalability
     

Learn More >

mipso-logo

 

Migration and Acceleration

  • Toolset delivering easy migration of existing AI applications
  • High-performance plug-and-play AI inference accelerator

 

Learn More >

deepai-logo

 

Edge AI Training

  • AI training at the edge on FPGA with a 10x performance/cost advantage vs GPUs
  • Support for Tensorflow, PyTorch, and Keras
     

Learn More >

uncanny-tile-white


License Plate Recognition

  • End-to-End LPR Solution
  • High efficiency, low latency and scalability
  • Open REST API for custom integration
     

Learn More >

Get Started

Get Started with Xilinx Vitis Video Analytics SDK