Optimize Vision Processing with AI Accelerator Card

Oct 23, 2019

IEI monthly news
Optimize Vision Processing with AI Accelerator Card

Optimize Vision Processing with AI Accelerator Card

Undoubtedly, the improvement of vision analytics technology brings numerous insights and benefits to the society.

Detecting faces and movements of pedestrians in public places for the purpose of people counting and public safety enhancement is a common vision analytics application.

In order to immediately process and analyze a large number of image files, and compare the analysis results, it is required to have a high-performance hardware device and advanced software technology to handle the task.

IEI releases the Mustang series AI accelerator cards which is integrated with Intel® Arria® 10 FPGA / Intel® Movidius™ VPU and support Intel® Open Visual Inference & Neural Network Optimization (OpenVINO™) toolkit to provide an excellent solution for optimizing vision processing.

Why Intel® Solution

  • Cross platform: easy to be implemented in the existing projects
  • Heterogeneous: integrate all Intel’s accelerators
  • Fast integration: easy to convert a trained model to Intel platform.
  • Compact size: small in size and scalable.
  • Power efficiency: FPGA with 40W TDP; VPUx8 with 25W TDP.
  • Better performance/$/W: Intel® solutions have better FPS/$/W, compared to other solutions
why intel solution for ai inference

What OpenVINO™ toolkit offers

Intel® Distribution of OpenVINO™ toolkit is based on convolutional neural networks (CNN), the toolkit extends workloads across multiple types of Intel® platforms and maximizes performance.

It can optimize pre-trained deep learning models such as Caffe, MXNET, and ONNX Tensorflow. The tool suite includes more than 20 pre-trained models, and supports 100+ public and custom models (includes Caffe*, MXNet, TensorFlow*, ONNX*, Kaldi*) for easier deployments across Intel® silicon products (CPU, GPU/Intel® Processor Graphics, FPGA, VPU).

openvino toolkit workflow

Advantages of using VPU-based accelerator card

  • Low power consumption
  • High scalability
  • Multi-Tasks
  • Compact size
  • Longevity 5 years
  • Medium & low resolution image( < 1080P)
Mustang-M2BM-MX2 acclerator card

Mustang-M2BM-MX2

  • Deep learning inference accelerating M.2 BM key card with 2 x Intel® Movidius™ Myriad™ X MA2485 VPU, M.2 interface 22 mm x 80 mm, RoHS

Specifications - Product News

- IEI Mustang mini AI accelerator cards introduction
[English][Chinese]


Mustang-MPCIE-MX2 accelerator card

Mustang-MPCIE-MX2

  • Compact size mini PCIe 30x50mm Two Intel® Movidius™ Myriad™ X VPU can execute Two topologies simultaneously.

Specifications - Product News

- IEI Mustang mini AI accelerator cards introduction
[English][Chinese]


Mustang-V100-MX4 accelerator card

Mustang-V100-MX4

  • Computing Accelerator Card with 4 x Intel® Movidius™ Myriad™ X MA2485 VPU, PCIe Gen 2 x 2 interface, RoHS

Specifications - Product News


Mustang-V100-MX8 Computing Accelerator Card

Mustang-V100-MX8

  • Computing Accelerator Card with 8 x Movidius Myriad X MA2485 VPU, PCIe gen2 x4 interface, RoHS

Specifications - Brochure
- IEI Deep Learning Inference Acceleration Card - Mustang V100 [English][Chinese]

Advantages of using FPGA-based accelerator card

  • Low latency (3ms in ResNet-50)
  • Continued improved performance
  • Industrial grade
  • 10-year longevity
  • High resolution image (>1080P)
Mustang-F100 FPGA card

Mustang-F100

  • PCIe FPGA Highest Performance Accelerator Card with Arria 10 1150GX support DDR4 2400Hz 8GB, PCIe Gen3 x8 interface

- Specifications - Brochure

- Mustang-F100-A10 OpenVINO™ Toolkit Installation Guide
[English][Chinese]

New Products

AFL3-W07A-AL | IP65 Panel PC

AFL3-W07A-AL | IP65 Panel PC

  • W7” light industrial panel PC
  • 12V DC input with lockable DC jack
  • Selectable AT/ATX power mode
  • On board IEEE802.3 at standard

- Specifications - Product News


AFL3-WXXA-AL | IP64 Panel PC

AFL3-WXXA-AL | IP64 Panel PC

  • W10.1”, 12.1” , W15.6”, 18.5” panel PC
  • 9 V ~ 30 V wide range DC input with lockable DC jack
  • Selectable AT/ATX power mode
  • Support PoE PD IEEE803.2 af/at/bt

- Specifications - Product News

- IEI Unboxing|Fixing your problem with IEI panel PC|AFL3-W15A-AL [English][Chinese]

- IEI’s Cost Effective and Flexible Power over Ethernet Panel PC|AFL3-W10A-AL[English][Chinese]


PulM-1G8T-BCM5720 | NIC for PUZZLE-IN001

PulM-1G8T-BCM5720 | NIC for PUZZLE-IN001

  • Broadcom BCM5720 based Network Interface Card with 8 x RJ-45 by 4 PCIe 2.0 x 2 Interface

- Specifications - Product News

News

Event News

- IEI Welcomes Your Visiting to MEDICA 2019 in Germany

Press Release

- Industrial Power over Ethernet Solution

- Hottest Cloud Based Products Certified by Amazon & Microsoft


Video

- IEI Case Study | A Good assistant for Quality Control - AOI Mechanical Controller
[English][Chinese]

- IEI Mustang mini AI accelerator cards introduction [English][Chinese]

- IEI Unboxing|Fixing your problem with IEI panel PC|AFL3-W15A-AL [English][Chinese]

- 2020 IEIxQNAP 高峰論壇:助您搭上數位轉型的列車![Chinese]


Case Study

- Virtual Machine Backups

Resource

 
catalog White Paper Case Study newsroom
Product Notice : PCN/EOL/TN/PDD