justadudewhohacks / face-api.js

JavaScript API for face detection and face recognition in the browser and nodejs with tensorflow.js
MIT License
16.7k stars 3.71k forks source link

face landmark detection model Architecture? #403

Closed Rakeshvcr closed 5 years ago

Rakeshvcr commented 5 years ago

What is the architecture of the face_landmark_68_model. I wanted to train my own landmark detection model with lesser points. please let me know the model architecture.

thirukumars commented 5 years ago

ssd model or tinyYolo or mtcnn we can use any of those three

On Fri, Sep 6, 2019 at 2:49 PM RAKESH_v notifications@github.com wrote:

What is the architecture of the face_landmark_68_model. I wanted to train my own landmark detection model with lesser points. please let me know the model architecture.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/justadudewhohacks/face-api.js/issues/403?email_source=notifications&email_token=AJOGHY3P43J52VXYTSDIGZ3QIIOCXA5CNFSM4IUG3F62YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HJYF42Q, or mute the thread https://github.com/notifications/unsubscribe-auth/AJOGHY7V5CYQJQG34FRC7NDQIIOCXANCNFSM4IUG3F6Q .

Rakeshvcr commented 5 years ago

@thirukumars correct me if I'm wrong, I think all those model are only for detecting faces, If the faces are detected then, they are using face_landmark_68_model to detect landmark.

justadudewhohacks commented 5 years ago

@thirukumars correct me if I'm wrong, I think all those model are only for detecting faces, If the faces are detected then, they are using face_landmark_68_model to detect landmark.

Correct.

From the README:

68 Point Face Landmark Detection Models This package implements a very lightweight and fast, yet accurate 68 point face landmark detector. The default model has a size of only 350kb (face_landmark_68_model) and the tiny model is only 80kb (face_landmark_68_tiny_model). Both models employ the ideas of depthwise separable convolutions as well as densely connected blocks. The models have been trained on a dataset of ~35k face images labeled with 68 face landmark points.

The implementation is a bit hidden, but you can find the code of the forward method here.

IGhadalharbi commented 5 years ago

hello please i have question ,i need to know how the function will take the landmark of face? and what is the landmark will take it? thank u very much

IGhadalharbi commented 5 years ago

hello why there is no answer? thank u