Open GeorgeS2019 opened 1 year ago
Thanks for the request, but I can't find any information about it in the suggested documentation below.
Latest Documentation for April 2023
Unfortunately, I can't understand what you're trying to tell with this post. Please note that I can't speak English. I'd appreciate if you could keep posts clear, simple and easy to understand.
Step1 :
face_blendshapes.tflite
Step2 : according to this project
- Converts .tflite into ONNX using tflite2onnx
- Replace Pad operators with combinations of ConstantOfShape and Concat.
- Add Expand operators to PRelu slope inputs.
Here is the unit test to work on the 52 blendshapes https://github.com/google/mediapipe/blob/master/mediapipe/tasks/cc/vision/face_landmarker/face_blendshapes_graph_test.cc
That's the link to the source code. Where can I download the file itself?
URL link to face_blendshapes.tflite
Ask in the issue, or you need to build it or WAIT
Ask in the issue, or you need to build it or WAIT
The .tflite
model file has to be converted from the TensorFlow model trained by the MediaPipe team, so only they can "build" the file.
Now I'm confused -- You suggested integrating it to this project, so I thought that you had it or you know where it was.
I'd like to close this issue temporarily if your suggestion is "WAIT". I'll reopen it when it gets ready.
Thanks for the update. Okay, it seems worth a try.
@keijiro
if you need to interact with others working on blendshape Here
The new Face Landmarker solution has been published, with support for ARKit Blendshapes. https://developers.google.com/mediapipe/solutions/vision/face_landmarker
By integrating the new face_blendshapes.tflite into your project, the generated 52 blendshapes coefficients could be used to drive the face of the 3D avatars from ReadyPlayer.ME.
Latest Documentation for April 2023