NCKU Aitum Angelfish

House Keeper - Smart Face Recognition System

Introduction

Our goal is to implements light face detection CNN model on ARC IoTdk board, edge devices such as ARC IoTdk board has advantages like low power comsumption and reasonable price are very suitable for preprocessing or wake-up device of the back-end system(in this case, we use face recognition CNN model as our backend system).

Demo Video

LINK

System architecture

Required Hardware

• Camera : OV7670

• Always-On Face Detector Device : ARC IoTDK

• Micro USB Cable

Hardware cable connection

1.Connect ov7670 to IoTDK Board.

ov7670	ARC: 2x18 Pin Extension Header
SCL	I2C0_SCL (need pull-up 10K resistor)
SDA	I2C0_SDA (need pull-up 10K resistor)

ov7670	ARC: Arduino PIN
VSYBC	arduino IO0
PCLK	arduino IO1
XCLK	arduino IO3
D7~D0	arduino IO4~IO11
3V3	+3.3V
RESET	+3.3V
GND	GND
PWDN	GND

2.Connect IoTDK Board to computer with micro USB Cable.

Required Software

• MetaWare Development Toolkit

• embARC Machine Learning Inference Library

• embARC Open Software Platform (OSP)

User Manual

Before running

• Download source code from github

• Setup hardware connection

• Goto (embarc_osp)\middleware\common\console_io.c

• Change

    console_uart = uart_get_dev(CONSOLE_UART_ID);
  
    console_uart->uart_open(BOARD_CONSOLE_UART_BAUD);
  
    xdev_in(console_getchar);
    xdev_out(console_putchar);
  

• to

    console_uart = uart_get_dev(CONSOLE_UART_ID);
  
    console_uart->uart_open(1500000);
  
    xdev_in(console_getchar);
    xdev_out(console_putchar);
  

• Goto python\read.py and python\build.py

  • In port='COM3',change COM3 to the port which your IoTDK Board connect.

Run the application

To start face detection and recognition

• Type python read.py and wait the terminal to print Configured

• Type make run to download and start the program on IoTDK Board

• If terminal shows memory allocation error, open obj_iotdk_10\mw_arcem9d\linker_mw.ldf

• change

      GROUP : {
      .mli_model_p2 ALIGN(8): {
      _f_mli_model_p2 = .;
      *(.mli_model_p2 .mli_model_p2.*)
      _e_mli_model_p2 = .;
      }
      } > REGION_DCCM AT > REGION_ICCM
  
      _load_addr_mli_model_p2 = LOADADDR(.mli_model_p2);
  
      GROUP : {
      .mli_model ALIGN(8): {
      _f_mli_model = .;
      *(.mli_model .mli_model.*)
      _e_mli_model = .;
      }
      } > REGION_XCCM AT > REGION_ICCM
  

• to

    GROUP : {
    .mli_model_p2 ALIGN(8): {
    _f_mli_model_p2 = .;
    *(.mli_model_p2 .mli_model_p2.*)
    _e_mli_model_p2 = .;
    }
    } > REGION_DCCM AT > REGION_DCCM
  
    _load_addr_mli_model_p2 = LOADADDR(.mli_model_p2);
  
    GROUP : {
    .mli_model ALIGN(8): {
    _f_mli_model = .;
    *(.mli_model .mli_model.*)
    _e_mli_model = .;
    }
    } > REGION_DCCM AT > REGION_DCCM
  

• Type make run to download and start the program on IoTDK Board again

To sign up a new face ID in face recognition database

• Type python build.py (yourname) ,you can set (yourname) parameter to any name you want and wait the terminal to print Configured.

• Type make run to download and start the program on IoTDK Board.

• Type any key and the system will sign up camera's current frame with (yourname).