Person and Face Detection using Intel OpenVINO toolkit

People and Face Detection in Video for Security Surveillance using the Intel Distribution of OpenVINO toolkit.

IntermediateProtip3 days2,422

Early Submission Prizes

Deep Learning Superhero Challenge

Person and Face Detection using Intel OpenVINO toolkit

Things used in this project

Software apps and online services

Intel OpenVINO™ toolkit

Python 3

Story

This project can be used to detect the presences of people in a room. It uses the pre-trained models downloaded using the OpenVINO model downloader.

Demonstration Video

This project can be used to detect the presence of people by detecting their faces and their body.

Steps

1 Download and install Anaconda Distribution of Python

2 Install all the dependencies

3 Download the Intel Distribution of OpenVINO toolkit.(My OpenVINO version - 2020.0.1.033)

Python Virtual Environment

1 Install virtual environment

pip install virtualenv

virtualenv --system-site-packages -p python openvinoenv

2 Activate virtual environment

openvinoenv\Scripts\activate

3 Running "setupvars.bat" file to set up the OpenVINO environment variables

cd "C:\Program Files (x86)\IntelSWTools\openvino_2020.1.033\bin\"setupvars.bat

4 Installing all the requirements from the "requirements.txt" file

cd "C:\Program Files (x86)\IntelSWTools\openvino_2020.1.033\deployment_tools\model_optimizer>"

pip install -r requirements.txt

5 Once done with OpenVINO, deactivate the environment

deactivate

NOTE - I have created a batch file named "openvinoenv.bat" to run the following steps in one singe line.

call "openvinoenv\Scripts\activate"call "C:\Program Files (x86)\IntelSWTools\openvino_2020.1.033\bin\setupvars.bat"

With a single line, the python virtual environment for OpenVINO toolkit can be created and the OpenVINO environment variables can be initialized.

openvinoenv.bat

Pre-trained Models

It uses the pre-trained models from the Intel OpenVino pre-trained models.

The following pre-trained models were downloaded using the model downloader

python "C:\Program Files (x86)\IntelSWTools\openvino\deployment_tools\open_model_zoo\tools\downloader\downloader.py"

face-detection-adas-0001 - link
person-detection-retail-0013 - link

Demonstration Videos used from the sample videos of intel-iot-devkit Github Repository.

System specifications used

Processor - Intel Core i7 9750H

Ram - 16 GB

OS - Windows 10 64 bit

Final Code Github Repository Link - here

Date of Download of Intel® Distribution of OpenVINO toolkit for Windows - January, 2020 (w_openvino_toolkit_p_2020.1.033.exe)

Email ID used- same as the one with the hackster.io

main.py

import cv2 

net2 = cv2.dnn.readNet('person-detection-retail-0013.bin', 'person-detection-retail-0013.xml')
net = cv2.dnn.readNet('face-detection-adas-0001.bin', 'face-detection-adas-0001.xml')

cap = cv2.VideoCapture("video1.mp4")

font = cv2.FONT_HERSHEY_SIMPLEX

while cv2.waitKey(1) < 0:
    hasFrame, frame = cap.read()
    frame = cv2.resize(frame,(1280,720),fx=0,fy=0, interpolation = cv2.INTER_CUBIC)
    if not hasFrame:
        break

    face_blob = cv2.dnn.blobFromImage(frame, size=(672, 384))
    net.setInput(face_blob)
    out_face = net.forward()

    for detection in out_face.reshape(-1, 7):
        confidence = float(detection[2])
        xmin = int(detection[3] * frame.shape[1])
        ymin = int(detection[4] * frame.shape[0])
        xmax = int(detection[5] * frame.shape[1])
        ymax = int(detection[6] * frame.shape[0])

        if confidence > 0.5:
            cv2.rectangle(frame, (xmin, ymin), (xmax, ymax), color=(0, 255, 0))
            cv2.putText(frame,'FACE',(xmin, (ymin-10)),font, 0.4,(0, 255, 255),1,cv2.LINE_AA) 

    person_blob = cv2.dnn.blobFromImage(frame, size=(672, 384))
    net2.setInput(person_blob)
    out_person = net2.forward()

    for detection in out_person.reshape(-1, 7):
        confidence2 = float(detection[2])
        xmin2 = int(detection[3] * frame.shape[1])
        ymin2 = int(detection[4] * frame.shape[0])
        xmax2 = int(detection[5] * frame.shape[1])
        ymax2 = int(detection[6] * frame.shape[0])

        if confidence2 > 0.6:
            cv2.rectangle(frame, (xmin2, ymin2), (xmax2, ymax2), color=(0, 255, 0))
            
            cv2.putText(frame,'PERSON',(xmin2, (ymin2-10)),font, 0.4,(0, 255, 255),1,cv2.LINE_AA) 
    cv2.imshow('OpenVINO face detection', frame)