Setting up a device service in an EdgeX instance

Necessary prerequisites

• have Docker and Docker-Compose installed
• have a make program, run this:

  $ sudo apt install build-essential

• have Golang installed and have your GOPATH set to ~/go. Check that with

  $ go env

  It is essential your device-sdk-go folder is inside your GOPATH directory.

Get the EdgeX device SDK and the device-goface files

Create folders to download SDK to:

$ mkdir -p ~/go/src/github.com/edgexfoundry
$ cd ~/go/src/github.com/edgexfoundry
$ git clone https://github.com/edgexfoundry/device-sdk-go.git

Clone the device-goface files from the project repo:

$ git clone https://github.com/oliviakumar/vmware_edgewalk

Change your directory in terminal to the device-goface folder (containig the makefile) and build the service

$ make build

If there are no error, your service should be ready to run. Change directories to cmd/device-goface and see if there's an executable "device-goface". If yes, execute it, if no, try to build again (in the higher device-goface), using

$ go build

Execute the built file:

$ ./device-goface

Watch the logs in the terminal to see if it works and check in Postman or browser: http://localhost:48080/api/v1/event/device/RandNum-Device-01/100

For any more help, consider the tutorial at: https://docs.edgexfoundry.org/Ch-GettingStartedSDK-Go.html

Goface

Necessary Prerequisites

To be able to run go-face, or programs that use go-face, on your local machine, follow
these steps on:

Kagami go-face github

Go-face Usage

To use go-face in your go program:

import "github.com/Kagami/go-face"  

Go-face Training and Inference

• Train()

  • Traverses through a fixed directory called "trainImages" and trains the model for every folder found
  • Adds the directory name, image data, and indices of the samples to the Models struct, which is later used for inferencing

• Infer()

  • Parameters:

    edgexcontext - edgex event instance
    params - parameters of the data interface that will be populated 

  • Returns:

    boolean - true or false depending on whether the face matches  
    interface - contains the data of matched person, or default values for anonymous  

  • Takes in the edgex event instance and the data interface from the app sdk and populates the data interface according to the results of the facial recognition
  • If face matched, returns true and populates the data interface according accordingly. If face not recognized,
    returns false and populates the data interface with default information

• TestForFace()

  • Parameters:

    imgPath - name of the image to check

  • Returns:

    boolean - true if given image path contains a face, false otherwise

  • Checks to see if the given image contains a face