Getting Started

In this tutorial, we shall go through the complete process of using DeepStack to build a Face Recognition system.

Setting Up DeepStack

Install and Setup DeepStack Using the Install Guide. If you have a system with Nvidia GPU, follow instruction on Using DeepStack with NVIDIA GPU to install the GPU Version of DeepStack

Starting DeepStack on Docker

Below we start DeepStack with only the face APIs enabled.

sudo docker run -e VISION-FACE=True -v localstorage:/datastore -p 80:5000 \
deepquestai/deepstack

sudo docker run --rm --runtime=nvidia -e VISION-FACE=True -v localstorage:/datastore \
-p 80:5000 deepquestai/deepstack:gpu

Basic Parameters

-e VISION-FACE=True This enables the face recognition APIs, all apis are disabled by default.

-v localstorage:/datastore This specifies the local volume where deepstack will store all data.

-p 80:5000 This makes deepstack accessible via port 80 of the machine.

Starting DeepStack on Raspberry PI

sudo deepstack start "VISION-FACE=True"

Starting DeepStack on Windows

Start the DeepStack App, Click Start Server, Select the Face API and click Start Now

Face Recognition

Think of a software that can identify known people by their names. Face Recognition does exactly that. Register a picture of a number of people and the system will be able to recognize them again anytime. Face Recognition is a two step process: The first is to register a known face and second is to recognize these faces in new pictures.

REGISTERING A FACE

Here we are building an application that can tell the names of a number of popular celebrities. First we collect pictures of a number of celebrities and we register them with deepstack

Below we will register the faces with their names

const request = require("request")
const fs = require("fs")
​
run_prediction("cruise.jpg","Tom Cruise")
run_prediction("elba.jpg","Idris Elba")
run_prediction("perri.jpg","Christina Perri")
run_prediction("adele.jpg","Adele")
​
function run_prediction(image_path,userid){
​
    image_stream = fs.createReadStream(image_path)
​
    var form = {"image":image_stream,"userid":userid}
​
    request.post({url:"http://localhost:80/v1/vision/face/register", formData:form},function(err,res,body){
​
        response = JSON.parse(body)
        console.log(response)
​
    })
​
}

Response

{ success: true, message: 'face added' }
{ success: true, message: 'face added' }
{ success: true, message: 'face added' }

RECOGNITION

const request = require("request")
const fs = require("fs")
​
image_stream = fs.createReadStream("test-image.jpg")
​
var form = {"image":image_stream}
​
request.post({url:"http://localhost:80/v1/vision/face/recognize", formData:form},function(err,res,body){
​
    response = JSON.parse(body)
    predictions = response["predictions"]
    for(var i =0; i < predictions.length; i++){
​
    console.log(predictions[i]["userid"])
​
    }
​
})

Response

Adele

We have just created a face recognition system. You can try with different people and test on different pictures of them.

The next tutorial is dedicated to the full power of the face recognition api as well as best practices to make the best out of it.

Performance

DeepStack offers three modes allowing you to tradeoff speed for peformance. During startup, you can specify performance mode to be , “High” , “Medium” and “Low”

The default mode is “Medium”

You can specify a different mode during startup as seen below as seen below

sudo docker run -e MODE=High VISION-FACE=True -v localstorage:/datastore -p 80:5000 \
deepquestai/deepstack

sudo docker run --rm --runtime=nvidia -e MODE=High -e VISION-FACE=True -v localstorage:/datastore \
-p 80:5000 deepquestai/deepstack:gpu

Note the -e MODE=High above

On Windows, you can easily select the High mode in the UI

Note the High radio button selected above

Speed Modes are not available on the Raspberry PI Version