Difference between revisions of "BPI-EAI80 AIoT 开发板"

From Banana Pi Wiki
Jump to: navigation, search
(→‎Software)
(→‎软件开发)
 
(9 intermediate revisions by the same user not shown)
Line 8: Line 8:
 
[[File:BPI-K210_1.JPG|thumb|[[BPI-K210 RISC-V AIoT board]]]]
 
[[File:BPI-K210_1.JPG|thumb|[[BPI-K210 RISC-V AIoT board]]]]
  
−
BPI-EAI80 AIoT board use Edgeless EAI80 chip design .it have Dual-Cortex M4F@200MHz 500DMIPS and AI-NPU：CNN-NPU @300 MHz 300GOPS. support LVDS pannel and camera interface. onboard wifi
+
BPI-EAI80 AIoT 开发板采用格力零边界EAI80芯片设计.支持Dual-Cortex M4F@200MHz 500DMIPS和AI-NPU：CNN-NPU @300 MHz 300GOPS. 支持LVDS屏接口和摄像头接口，板载wifi,方便的做各种AI物联网应用。
 +
 
 +
开发板设有CAN总线、双麦克风、单摄像头、USB 2.0 Type-C，1024*768分辨率的触控TFT-LCD接口，支持低功耗唤醒和电池备份域，最大提供8MB SDRAM + 32MB SPI Nor Flash的组合，以及提供常规的调试、中断和复位。配合ESP8266，可以实现网络资源获取、远程操控和OTA升级。EAI80内置高性能多媒体模块，图像专用传输通道和包括混叠、融合、缩放、格式转换、角点检测等预处理功能，音频支持最高8路的PDM/I2S麦克风，芯片其他功能可以通过GPIO口的复用和模拟实现。
 +
用户和开发者可以充分利用MCU的控制、NPU的加速和CAN总线的交互，为工业应用加入音视频的前处理和识别能力，为产品赋能和提速。
  
 
==目标应用==
 
==目标应用==
  
−
*Voice Control - Key words real-time control
+
*语音控制-关键字实时控制
−
*Computer Vision - Detection & recognition of object and biology (Face, Body, Gesture), vSLAM
+
*计算机视觉-物体和生物(脸，身体，姿势)的检测和识别，vSLAM
−
*End-side AIoT - Edge computing, Info. Security,Off-line equipment control, System monitor
+
*终端AIoT -边缘计算，信息。安全，离线设备控制，系统监控
−
*Sensor, Attendance machine, AD display, Wearable,devices, Smart unmanned retail, STEM education
+
*传感器、考勤机、广告显示、可穿戴设备、智能无人零售
−
*Home & Building Automation – White goods,HVAC, Lighting, Security system, IoT gateways
+
*家庭和建筑自动化-白色家电，暖通空调，照明，安全系统，物联网网关
−
*Industrial Computing - EBS, PLCs, M2M, T&M,Auto-factory, HMI control assembly, QR/bar code
+
*工业计算- EBS, plc, M2M, T&M,Auto-factory, HMI控制总成，二维码
−
*Motor Control & Power Conversion - VFC,FOC, 3D/thermal Printers, ADAS, UAV, Robots
+
*电机控制和功率转换- VFC,FOC, 3D/热敏打印机，ADAS，无人机，机器人
−
*STEAM education
+
*STEM教育
  
 
=硬件=
 
=硬件=
Line 39: Line 42:
 
[[File:Banana_PI_BPI-EAI80_interface_750.JPG]]
 
[[File:Banana_PI_BPI-EAI80_interface_750.JPG]]
  
−
=软件=
+
=github源代码=
 +
 
 +
*github: https://github.com/BPI-SINOVOIP/BPI-EAI80-bsp
 +
 
 +
=软件开发=
 +
 
 +
==BPI-EAI80文档==
 +
 
 +
*BPI-EAI80 原理图: https://pan.baidu.com/s/13L-rSELVP3XisHmm6c251Q 提取码: 6qb8
 +
*BPI-EAI80 DXF 结构文件：https://pan.baidu.com/s/13L-rSELVP3XisHmm6c251Q 提取码: 6qb8
 +
 
 +
==Linux SDK Compile and burn ==
 +
 
 +
*1. [[EAI80 Burning Steps (Elementary)]]
 +
 
 +
==BPI-EAI80 board , EAISeries SDK UserGride==
 +
 
 +
BPI-EAI80 board  , EAISeries SDK UserGride:
 +
 
 +
https://drive.google.com/file/d/18kmhdspeWu05nkt3DMM2rAd1oz-30rQR/view?usp=sharing
 +
 
 +
please install KeliSDK:
 +
 
 +
https://github.com/BPI-SINOVOIP/BPI-EAI80-bsp/tree/master/KelisSDK
 +
 
 +
https://github.com/BPI-SINOVOIP/BPI-EAI80-bsp/blob/master/Keil.EAISeries_DFP.1.4.1.pack
 +
 
 +
windows version:
 +
 
 +
https://github.com/BPI-SINOVOIP/BPI-EAI80-bsp/tree/master/KelisSDK
 +
 
 +
 
 +
1. Install EAI80 SDK (windows )
 +
2. PLease install ※Keil.EAISeries_DFP.1.4.1.pack§first;
 +
3. Demo board should connect with PC through J-Link and SWD interface. All the operations refer to "EAISeries_SDK_UserGuide" in detail.
 +
 
 +
EAI80 Demo:
 +
 
 +
Demo path: EAI80_SDK_v1.0\ugelis\kelis_example\ai_example\ai
 +
Demo functions:
 +
* 1st.Voice - Key words recognition;
 +
* 2rd.Computer Vision - Hand gesture detection and recognition;
 +
* 3nd.Computer Vision - Human dody detection.
 +
 
 +
=关于格力零边界EAI系列芯片=
 +
 
 +
格力零边界EAI系列芯片采用“双cpu +NPU”结构，将人工智能能力与实时低功耗嵌入式单片机相结合。丰富的外设和hw安全集成，EAI芯片系列是为支持下一代AIoT产品而设计的物联网应用，提供带有人工智能识别和单片机控制的单片机交钥匙解决方案。格力零边界EAI系列嵌入式AI系列交叉处理器.在AIoT的浪潮下，面向智能家居、工业、stem教育、能源管理等领域推出，具备AI能力，低功耗，高性价比，提供完整的HW/SW交钥匙解决方案。目标建立世界上最小的AI+MCU模块, 促进跨界创新，增强终端和产业的权能。.
 +
* 更先进的结构和性能
 +
* 运行和待机功耗更低
 +
* 语音/计算机视觉/二维图形加速器
 +
* 实时操作/多硬件级安全
 +
* 易于使用/集成更丰富/成本更低
  
−
=About Edgeless EAI series=
+
EAI系列跨界AI MCU, CPU核心基于ARM Cortex-M4, ARMv7-M支持预定义的32位地址空间，对代码、数据和外设进行细分，对片内和片外资源进行细分，其中片内指与处理器紧密耦合的资源。
  
−
EAI Series adopt“Dual-CPU+NPU”structure, combine AI ability with real-time and low power embedded MCU. Rich peripherals and HW-security integrated, EAI series are designed to support next-gen. IoT applications,
+
EAI是一个多核微控制器，实现了双核Cortex-M4核。所有核心都可以访问完整的内存映射。ARM Cortex-M4用作主处理器。另一个ARM Cortex-M4内核可以作为协处理器来协助ARM Cortex-M4并执行复杂的数学计算。
−
provide single-chip turnkey solution with AI recognition and MCU control. Edgeless Embedded AI Series crossover processor, EAI series,under the wave of AIoT, are launched for smart home, industry, stem education, energy MGT and etc., with AI ability, low power and cost-effective, provide complete HW/SW turnkey solutions. Target to establish the smallest AI+MCU module in the world, promote transboundary innovation, empower end-sides and industries.
 
−
* More advanced structure and performance
 
−
* Lower run and standby power consumption
 
−
* Voice / Computer vision / 2D Graph accelerator
 
−
* Real-time operation / Multiple HW-level security
 
−
* Easy to use / Richer integration / Lower cost
 
  
−
EAI series crossover AI MCU, CPU core is based on ARM Cortex-M4, ARMv7-M supports a
+
EAI集成CNN处理器，使用深度学习方法处理图像检测和识别，性能高，能耗低。它支持主流CNN模式，如Resnet-18, Resnet-34, Vgg16, GoogleNet, Lenet等，卷积与内核大小从1到7，频道/特征数高达512，最大/平均池功能与内核
−
predefined 32-bit address space, with subdivision for code, data, and peripherals, and regions for
 
−
on-chip and off-chip resources, where on-chip refers to resources that are tightly coupled to the
 
−
processor.
 
−
EAI is a multi-core microcontroller implementing Dual-ARM Cortex-M4 cores. All cores have
 
−
access to the complete memory map. One ARM Cortex-M4 is used as the master processor. The
 
−
other ARM Cortex-M4 core can be used as a co-processor to off-load the ARM Cortex-M4 and to
 
−
perform complicated mathematical calculations.
 
−
CNN processor is integrated in EAI, which can handle image detection and recognition use deep
 
−
learning methods with high performance and low energy consumption. It supports mainstream
 
−
CNN model such as Resnet-18, Resnet-34, Vgg16, GoogleNet, Lenet etc, convolutional with kernel
 
−
size from 1 up to 7, channel/feature number up to 512, max/average pooling function with kernel
 
  
−
EAI chip Device Summary
+
EAI芯片设备概述
  
 
[[File:Eai80_chip.png]]
 
[[File:Eai80_chip.png]]
Line 72: Line 109:
 
* Edgeless EAI series datasheet: https://drive.google.com/file/d/1KSOOAXKe0eLpXGxulQPzuoHqoMPqTq99/view?usp=sharing
 
* Edgeless EAI series datasheet: https://drive.google.com/file/d/1KSOOAXKe0eLpXGxulQPzuoHqoMPqTq99/view?usp=sharing
 
* Edgeless EAI series reference manual : https://drive.google.com/file/d/1oviJX3j_HNM-kNA8Ceszbke5KXs3yXua/view?usp=sharing
 
* Edgeless EAI series reference manual : https://drive.google.com/file/d/1oviJX3j_HNM-kNA8Ceszbke5KXs3yXua/view?usp=sharing
 +
* Edgeless EAI series uGelis reference manual :https://drive.google.com/file/d/18rZth2K3mjTjVxeSC-CQBYwuLXqfI06d/view?usp=drivesdk
 +
* Edgeless EAI uGelis Ubuntu SDK Environments Establishment : https://drive.google.com/file/d/1ua8qQL3Ebqfy7DRSyUPVjw0x7US02Y1W/view?usp=drivesdk
 +
* BPI-EAI80 ESP8266 module default FW, if you want use this file ,you can burn it to ESP8266 by youself：https://drive.google.com/file/d/1wNCrayg0LcsNWinq1eGLdvr_XdTdMT9e/view?usp=drivesdk

Latest revision as of 17:07, 28 December 2020


BPI-EAI80 AIoT开发板介绍

BPI-EAI80 AI board Gree EAI80 chip design
BPI-EAI80 AI board Gree EAI80 chip design
BPI-AI Kendryte K210 RISC-V
BPI-Bit with ESP32 design
BPI-K210 RISC-V AIoT board

BPI-EAI80 AIoT 开发板采用格力零边界EAI80芯片设计.支持Dual-Cortex M4F@200MHz 500DMIPS和AI-NPU：CNN-NPU @300 MHz 300GOPS. 支持LVDS屏接口和摄像头接口，板载wifi,方便的做各种AI物联网应用。

开发板设有CAN总线、双麦克风、单摄像头、USB 2.0 Type-C，1024*768分辨率的触控TFT-LCD接口，支持低功耗唤醒和电池备份域，最大提供8MB SDRAM + 32MB SPI Nor Flash的组合，以及提供常规的调试、中断和复位。配合ESP8266，可以实现网络资源获取、远程操控和OTA升级。EAI80内置高性能多媒体模块，图像专用传输通道和包括混叠、融合、缩放、格式转换、角点检测等预处理功能，音频支持最高8路的PDM/I2S麦克风，芯片其他功能可以通过GPIO口的复用和模拟实现。 用户和开发者可以充分利用MCU的控制、NPU的加速和CAN总线的交互，为工业应用加入音视频的前处理和识别能力，为产品赋能和提速。

目标应用

  • 语音控制-关键字实时控制
  • 计算机视觉-物体和生物(脸，身体，姿势)的检测和识别，vSLAM
  • 终端AIoT -边缘计算，信息。安全，离线设备控制，系统监控
  • 传感器、考勤机、广告显示、可穿戴设备、智能无人零售
  • 家庭和建筑自动化-白色家电，暖通空调，照明，安全系统，物联网网关
  • 工业计算- EBS, plc, M2M, T&M,Auto-factory, HMI控制总成，二维码
  • 电机控制和功率转换- VFC,FOC, 3D/热敏打印机，ADAS，无人机，机器人
  • STEM教育

硬件

硬件规格

  • CPU Dual-Cortex M4F@200MHz 500DMIPS
  • AI-NPU：CNN-NPU @300 MHz 300GOPS
  • 2D Graph :Dual-Camera Max
  • SDRAM 8M
  • LCD 1024*768 TFT-LCD
  • CANBUS 2.0 A/B
  • ESP8266 Wifi onboard
  • 40PIN GPIO (share with LCD )
  • 2 Mic support
  • Size: 86x54mm

硬件接口示意

Banana PI BPI-EAI80 interface 750.JPG

github源代码

软件开发

BPI-EAI80文档

Linux SDK Compile and burn

BPI-EAI80 board , EAISeries SDK UserGride

BPI-EAI80 board , EAISeries SDK UserGride:

https://drive.google.com/file/d/18kmhdspeWu05nkt3DMM2rAd1oz-30rQR/view?usp=sharing

please install KeliSDK:

https://github.com/BPI-SINOVOIP/BPI-EAI80-bsp/tree/master/KelisSDK

https://github.com/BPI-SINOVOIP/BPI-EAI80-bsp/blob/master/Keil.EAISeries_DFP.1.4.1.pack

windows version:

https://github.com/BPI-SINOVOIP/BPI-EAI80-bsp/tree/master/KelisSDK


1. Install EAI80 SDK (windows ) 2. PLease install ※Keil.EAISeries_DFP.1.4.1.pack§first; 3. Demo board should connect with PC through J-Link and SWD interface. All the operations refer to "EAISeries_SDK_UserGuide" in detail.

EAI80 Demo:

Demo path: EAI80_SDK_v1.0\ugelis\kelis_example\ai_example\ai Demo functions:

  • 1st.Voice - Key words recognition;
  • 2rd.Computer Vision - Hand gesture detection and recognition;
  • 3nd.Computer Vision - Human dody detection.

关于格力零边界EAI系列芯片

格力零边界EAI系列芯片采用“双cpu +NPU”结构，将人工智能能力与实时低功耗嵌入式单片机相结合。丰富的外设和hw安全集成，EAI芯片系列是为支持下一代AIoT产品而设计的物联网应用，提供带有人工智能识别和单片机控制的单片机交钥匙解决方案。格力零边界EAI系列嵌入式AI系列交叉处理器.在AIoT的浪潮下，面向智能家居、工业、stem教育、能源管理等领域推出，具备AI能力，低功耗，高性价比，提供完整的HW/SW交钥匙解决方案。目标建立世界上最小的AI+MCU模块, 促进跨界创新，增强终端和产业的权能。.

  • 更先进的结构和性能
  • 运行和待机功耗更低
  • 语音/计算机视觉/二维图形加速器
  • 实时操作/多硬件级安全
  • 易于使用/集成更丰富/成本更低

EAI系列跨界AI MCU, CPU核心基于ARM Cortex-M4, ARMv7-M支持预定义的32位地址空间，对代码、数据和外设进行细分，对片内和片外资源进行细分，其中片内指与处理器紧密耦合的资源。

EAI是一个多核微控制器，实现了双核Cortex-M4核。所有核心都可以访问完整的内存映射。ARM Cortex-M4用作主处理器。另一个ARM Cortex-M4内核可以作为协处理器来协助ARM Cortex-M4并执行复杂的数学计算。

EAI集成CNN处理器，使用深度学习方法处理图像检测和识别，性能高，能耗低。它支持主流CNN模式，如Resnet-18, Resnet-34, Vgg16, GoogleNet, Lenet等，卷积与内核大小从1到7，频道/特征数高达512，最大/平均池功能与内核

EAI芯片设备概述

Eai80 chip.png

EAI series Documents

Retrieved from "https://wiki.banana-pi.org/index.php?title=BPI-EAI80_AIoT_开发板&oldid=10088"