Skip to content

Make Baidu EdgeBoard Lite as a general Zynq FPGA development board

License

Notifications You must be signed in to change notification settings

changhai0109/edgeboard

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

logo

Awesome EdgeBoard Lite

Use EdgeBoard Lite as a general Zynq development board

en

EdgeBoard Lite (a.k.a. FZ3) is an FPGA-accelerated embedded AI computing board, launched by Baidu Brain in 2019. As part of the Baidu AI solution, it can quickly depoly the neural network models.

The board is based on Xilinx Zynq Ultrascale+ ZCU3EG MPSoC, which can achieve 1~2 TOP/s with an average power of 5~10 W. It also has the following peripherals:

  • 2GB DDR4 RAM
  • 8GB eMMC Flash
  • 256Mb QSPI Flash
  • PCIe x1 connector
  • MIPI camera connector
  • BT1120 HDTV connector
  • 1000Mbps Ethernet
  • TF-card slot
  • USB Hub (1 x USB 3.0, 1 x USB 2.0)
  • Mini DisplayPort (miniDP)
  • USB UART

In addition to the rich peripherals, what matters more is that it only costs ~1000 CNY (about 140 USD)! It is almost the cheapest Xilinx Zynq Ultrascale+ development board. Another Zynq ZCU3EG board Ultra96, much loved by developers, is priced at 249 USD. In the Chinese market, Ultra96 even costs more than 2200 CNY, more than double of EdgeBoard Lite.

Unfortunately, Baidu sells the EdgeBoard as its AI solution instead of a general-purpose development board. Therefore, it is a bit difficult for us to obtain full official support (You can send emails to Baidu to request limited documents).

This repository contains materials that help us use EdgeBoard as a general Ultrascale+ Zynq development board.

Documents

Here are some documents you may need, and you can find them in the ./docs folder. It should be noted that most of them are in Chinese, and currently do not have an English translation.

Most of above documents are from ALINX, the manufacturer of this board.

Documents about how to run EasyDL/PaddlePaddle models on the EdgeBoard Lite will not be listed here. You can visit its product website to access them.

Board Files & PYNQ Image

I have ported the PYNQ framework to EdgeBoard Lite, and provide the prebuilt image file. All the necessary source code are also open sourced.

Vivado board files

They are in ./board_files. Please read its inside README for more details.

PYNQ image file

The prebuilt image file is compiled based on PYNQ v2.7. Because it is too large (~10GB, exceed the GitHub limit), I put it on Aliyun Cloud Drive.

The folder ./pynq is board specification files, which are necessary for PYNQ compilation. If you want to build from scratch, you can read this post on my blog (sorry that it is in Chinese).

License

Except for those already noted in the text, other documents written by myself are released under the CC-BY 4.0 and the code is released under the MIT license.

This repository is created and maintained by me personally and has nothing to do with Baidu.

About

Make Baidu EdgeBoard Lite as a general Zynq FPGA development board

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 58.6%
  • Shell 41.4%