I want to create full face masks based on the 468 landmarks from the sample face mesh scene.
This article contains the steps for creating a face mask based on mediapipe's face_geometry obj face mask model: https://qiita.com/yamatohkd/items/b2d9a6055761b0b2c369
It's in Japanese, just Translate the page to English.
It is very simple.
We just get the normalized values(NormalizedLandmarkList) from the Draw Method.
After that, we apply the values obtained from there to the mesh vertices of the obj file above.
The face_geometry obj model can be downloaded from here: https://github.com/google/mediapipe/blob/master/mediapipe/modules/face_geometry/data/canonical_face_model.obj
Are there better solutions?
In general, how should I go about solving this problem?
I tried some solutions by merging dlib algorithms with Mediapipe, but the face mask is extremely unstable. It keeps moving along with the camera. I need a production ready solution for a commercial app.
Description
Hi,
I want to create full face masks based on the 468 landmarks from the sample face mesh scene. This article contains the steps for creating a face mask based on mediapipe's face_geometry obj face mask model: https://qiita.com/yamatohkd/items/b2d9a6055761b0b2c369 It's in Japanese, just Translate the page to English. It is very simple. We just get the normalized values(NormalizedLandmarkList) from the Draw Method. After that, we apply the values obtained from there to the mesh vertices of the obj file above. The face_geometry obj model can be downloaded from here: https://github.com/google/mediapipe/blob/master/mediapipe/modules/face_geometry/data/canonical_face_model.obj
Are there better solutions? In general, how should I go about solving this problem? I tried some solutions by merging dlib algorithms with Mediapipe, but the face mask is extremely unstable. It keeps moving along with the camera. I need a production ready solution for a commercial app.
Thanks.