Face detection Using Matlab Full Code|Image labeling|How to Train Object detector to detect Face.
In this Video, I have explained
1. How Face detection Works
2.How to Train ACF Object Detector to Detect Faces
3.How AI learns to detect Face.
4.Coding Part Simple and Easy
Code Link :-
https://drive.google.com/file/d/1zXRj...
Website Link :-https://facedetectionmatlabacfobjectd...
Time stamps
Demo of how accurate it detects Faces in an Image
1:09
How AI detects Faces ?
2:50
image labeller App in Matlab
3:25
image labelling
4:25
What kind of Face images are easy to train
4:50
Coding
7:35
Acf object detector
12:20
Testing on Image
17:30
Testing on Image
18:10
Testing on Video
19:40
How Face detection Works ?
Human Face consists of Various different Features like Shape of Mouth,Eye,Eyebrow,Nose,Moutache,Skin Color,HOG etc.
Now, An AI Extract this features and learns Over very large number of Dataset.
Labelling , Training & Testing.
How to Train an AI to detect any object or face?
1. Take large Dataset of face i.e Large number of images having face.
2.Labelling :- Label the Face region in an Image, by Drawing Rectangle i.e Bounding Boxes.
3.The Bounding boxes region is indicating face region, So AI will extract
features of that Region.
4. Give Negative images to AI i.e Images not having Face.
5. So posiive Images and Negative images given to AI, So it will learn to differeniate.
6.Run Object Detector.
7.Train Object detector to Certain Stages, So it can Learn to detect faces.
8.Once Detector is Trained to detect faces.
9.Now, U can Test Detector on any unlabellled i.e images on which detector is not trained , also It will work on detecting Face in Video i.e Face Tracking.
Now , Tips for Labelling
1. Now , U have to make sure that the Face images that u are using should be not processed like if it is Filtered,animated than it, becomes tough to Train .
2. So CAmera images Recommended , U can also Download FACE dataset images.
3. images with Different Backrounds, Side faces, Frontal face, different lighting conditions etc. should be taken.