how to get the facial feature points from the faces???

Using the Viola Jones algorithm m able to detect the face parts. Now i want to extract the facial feature points from eyes,nose and mouth. how do i do that?? which method to use?? which function to use?? can anyone plz give me some suggestions..

[Hawa Mohammad]

have you already get the answer? can you help me please? i need to find the 20 features points but still cannot find the right code.

[Mohamed]

can you share code to find 20 feature points

[fatin nabila othman]

Hi can you share with me the codes?

[Sidra Ashraf]

Hi can you please share the code.

[Pavithra]

hi can you please share with me the code?

Answer 1

[Image Analyst]

I'm sure there are lots of published methods, and I don't work in the face recognition field, so you're best off looking up published papers of those who do. Look here for an algorithm that seems promising: Recognition, Detection, Tracking, Gesture Recognition, Fingerprints, Biometrics

[Isha Pandya]

Can LDA algorithm perform facial feature selection and extraction?

[Image Analyst]

I don't see how. How are you envisioning it to be able to do that? What would be your input (the entire image???) and what features would be the output?

[Isha Pandya]

I have referred some of the published papers for facial feature extraction. They mentioned LDA is a good method for dimentionality reduction/ feature extraction. I have previously detected faces from images. Then I cropped the images to get full frontal view of the face. Now I want to extract facial features like eyes, Eyebrows and lips. And this should be done using LDA. But I am facing difficulty to do so. Can you please help me with the code?

[Image Analyst]

Are their images color or monochrome? Are they using LDA on the original image or on the feature vectors?

Answer 2

[Anand]

You can feed the bounding box of the detected face to any one of the feature detectors in the Computer Vision System Toolbox: detectFASTFeatures, detectHarrisFeatures, detectMinEigenFeatures, detectMSERFeatures or detectSURFFeatures. Here is an example of how to do this:

I  = imread('visionteam.jpg');

% Detect faces
faceDetector = vision.CascadeObjectDetector;
bboxes = step(faceDetector, I);

% Select the first face
face = I(bboxes(1,2):bboxes(1,2)+bboxes(1,4),bboxes(1,1):bboxes(1,1)+bboxes(1,3));

% Detect SURF features
ftrs = detectSURFFeatures(face);

%Plot facial features.
imshow(face);hold on; plot(ftrs);

If you specifically want features around lets say the eyes, you can change the classification model to eyes and follow the same steps:

eyeDetector = vision.CascadeObjectDetector('eyepairsmall');
bboxes = step(eyeDetector,I);

[danny choy]

can combine more than one parts ? like i want eyes and nose

[Anand]

you can use two detectors for that:

noseDetector = vision.CascadeObjectDetector('nose');
bboxesNose = step(noseDetector,I);

[danny choy]

i tried two detectors but keep error. Can you show me how to use two detector ?

[Hawa Mohammad]

sir, i have used your code but seems like the points is scatter. how to get only 20 points on the facial features? could you please help me with the code?

[maryam asadzadeh]

Hi, did you find the matlab code for point detection?

[lê linh]

How can I compare points on r face to give database identity or dryness

Answer 3

[sudeep g]

I  = imread('visionteam.jpg');
% Detect faces
faceDetector = vision.CascadeObjectDetector;
bboxes = step(faceDetector, I);
% Select the first face
face = I(bboxes(1,2):bboxes(1,2)+bboxes(1,4),bboxes(1,1):bboxes(1,1)+bboxes(1,3));
% Detect SURF features
ftrs = detectSURFFeatures(face);
%Plot facial features.
imshow(face);hold on; plot(ftrs);

[Fawad Chandio]

I don't want to use image you specifically read'visionteam.jpg'. I want use my browse image how can I do...?

[M Lorentschk]

with uigetfile: you can also specify a location where the browse window will start (here "dirPath")

dirPath = 'D:\example\data';
I = uigetfile(dirPath);

[Alawi Al-Saggaf]

How to compare two features for example ftrs1 of face1 and ftrs2 for face2.

What is the type of this ftr, it give me

3×1 SURFPoints array with properties:

Scale: [3×1 single]

SignOfLaplacian: [3×1 int8]

Orientation: [3×1 single]

Location: [3×2 single]

Metric: [3×1 single]

Count: 3

[Sakshi]

@sudeep g Thank you so much for this code. They work fine for me, but I want to ask if one can modify the code so that it will detect each point and then provide the coordinates of each feature. My aim is to find the vertical and horizontal distances of the face (working on an animal's face), and the mean value needs to be measured. Can you help me provide this code?