Let's Talk
Global (English)Global
Global (English)Global (English) Global (English) Global (English)United States (English) Global (English)Japan (Japanese)

Overview

VVDN, together with AMD and Kinara, offers VVDN-EAIA an edge AI appliance optimized for high-quality, real time video analytics for smart retail, smart cities and industrial use cases.

VVDN-EAIA includes a power-efficient, compact AMD Xilinx Kria K26 SOM integrated with 4 Kinara Ara-1 AI processors to support advanced AI analytics on up to 8 parallel video streams with a resolution of 1080P@30fps.

The edge appliance comes with a PetaLinux embedded SDK and Kinara SDK, which can seamlessly port user’s AI models and customize video pipelines. It is tailored to provide a solution that is out of the box ready for deployment with minimal FPGA programming required.

Features

  • Ability to handle 8 concurrent 1080p@30fps video streams
  • End-to-End hardware acceleration of AI inference models enabling real-time analytics for the 8 parallel streams
  • Standard vision pre- and post- processing functions automatically mapped to Kria’s programmable logic
  • Software and higher-level user applications run on AMD KRIATM k26 SOM
  • Kinara Ara-1 AI inference processor, with its latency optimized design, runs AI inference at a high performance, low latency and low power
  • Support for state-of-the-art AI models including vision transformers
  • Performance: Resnet-50: 400 IPS

Technical Specifications

SOM
  • AMD KRIATM K26 SOM
GPU
  • Arm Mali™-400 MP2 based graphics processing unit with an FMAX = 600 MHz
CPU
  • Arm® Cortex®-A53 based application processing unit (APU) consisting of quad-core Cortex-A53 processors
Vision Accelerator
  • Kinara Ara-1
  • 1GB LPDDR 4 (additional accelerator memory)
Camera
  • Upto x8 RTSP cameras through external router
Memory
  • 4GB DDR4 & 512Mb QSPI
USB
  • 3x USB 3.0
Network Connectivity
  • 2x GBE LAN, supported by ethernet controllers
Storage
  • 16GB eMMC on SOM memory
  • SSD up to 2TB (optional)
Video Encode
  • 4K @ 60FPS | 8x 1080p @ 30 (H.265/H.264)
Video Decode
  • 4K @ 60FPS | 8x1080p @ 30fps
Display
  • 1xDP 1.2a and 1xHDMI 1.4
Power Supply
  • DC supply, Adapter 12V@8.5A
Dimensions
  • 255.2mm x 205mm x 42mm
Other Interfaces
  • Alarm out, Status LED

TARGETED APPLICATIONS

QCS 410 SoM's compact size and features make it the best to facilitate a cost-effective and efficient system in which next-gen camera modules can be interfaced for various AI-based applications:

SECURITY SURVEILLANCE

SMART RETAIL

SMART CITY

Machine Vision

Robotics

Smart Factory

BOX CONTENTS

VVDN Edge AI Appliance Unit

Power Adaptor