Labels

 Output Video :-



Step 1 :- Image Labeling using Image labeler app


Drawing Rectangle into the Face mask Region



Steps :-

1.Face mask images of our own with different Background,Side Face,Frontal Face and Also you can Download Face Mask Images from the Internet but make sure they look Original not much Processed.

You can also Download FACE  Mask images for Free :-

300 images :-  Download Here




2. Labelling in Image labeler App.
Labelling means:- Drawing Rectangle Bounding Box around Face.

3.Exporting Labeled Images
 
4.Training ACF object Detector to Certain Stages.

5.Testing our AI on Unlabelled Images.

 

Implementation

Step 1:- Collect Sample Face images either Face dataset available on internet or Of your Own Images with different Backrounds and Light Conditions also.

Around  200-300 Faces MAsk images  are more than Enough for  Training Our Detector.

Warning:- Dont Use Images with same Background and Filtered Images from Insta or any such photo edited app.


Step 2:-  Open MATLAB 

Open Image labeler App by Clicking on APP icon and searching for image labeler.







Note :-Image labeler app  is available on Latest Versions of  MATLAB like 2017,2018,2019,2020 MATLAB Versions.

By Drawing Rectangle box i.e Boundary or Bounded Box on Face region of Image.

Now, After Labeling ,The AI extracts Features from the Bounding Box like Shape of Nose,Eyes,Mouth & Skin color.

We also need to supply Negative Images i.e images not having face,
so AI can learn to find Difference what a FACE is and What features face doesnt have.

But,  We are using ACF Object detector which doesnt Requires Negative Images, Only Positive Image with Labeling Face region are enough.
ACF Detector Creates its own Negative examples to Train.

Step 3:- Export Labeled images &
Load it in MATLAB, By Writing Command.

load('face.mat');

Step 4 :- Make Training Data folder ,
use Command to Train ACF DETECTOR..

CODE is Given in Video and also below .

ACF detector requires  10-20 Stages and it takes around 
10-20 minutes .

Once Detector is detected ,u can use it to detect faces in images as well as In Video


FACE MASK Tracking CODE:- 

LINK :-  https://drive.google.com/file/d/14cno2U3ccGUBAvjnylB_CNbFbvRAa8ze/view?usp=sharing


 load('Mask.mat');

 %  1. Load labelled data file ,which created through image labelling

FaceMask = selectLabels(gTruth,'Mask');

% 2. Creating Variable FaceGtruth in which will store labels 'Face'

if isfolder(fullfile('TrainingData'))
cd TrainingData
else 
mkdir TrainingData
end
addpath('TrainingData');

% 3. if else condition, if means 'if full name TrainingData exist ,locate that file ' else here if TrainingData Files doesnt exits Make one and 
%        add it to the MATLAB Path

trainingData = objectDetectorTrainingData(Faceamsk,'SamplingFactor',1,'writeLocation','TrainingData');
detector = trainACFObjectDetector(trainingData,'NumStages',20);
save('Detector.mat','detector');

rmpath('TrainingData');

% Make variable trainingData in which will store and passing Parameters like FaceGtruth that is labels , Sampling factor means Examples Face images
% if sampling factor is 2 than 2times negative images taken, Writing location = as TrainingData Folder


load('Detector.mat');

% load Trained Object Detector

vidReader = VideoReader('C:\Users\ANKIT TIWARI\Desktop\Face detection UTUBE\face Rawdata\faceMask.mp4');

% VideoReader is for reading Video

vidPlayer = vision.DeployableVideoPlayer;

% Play Video in MATLAB
i = 1;

results = struct('Boxes',[],'Scores',[]);

% results is variable for storing bounding box i.e rectangle values

while(hasFrame(vidReader))
% While loop  
I = readFrame(vidReader);
[bboxes, scores] = detect(detector,I,'Threshold',1);
[~,idx] = max(scores);
results(i).Boxes = bboxes;
results(i).Scores = scores;
annotation = sprintf('Mask' ,detector.ModelName,scores(idx));
I = insertObjectAnnotation(I,'rectangle',bboxes(idx,:),annotation);
step(vidPlayer,I);
i =i+1;
end
results = struct2table(results);


Labels:

Comments

Post a Comment