marekkowalski / holoface Goto Github PK

Couple of questions here..
Looking at this project it seems that it is over complicated, for example, when calculating the rotation vector, why not just use opencv's rodrigues function? also why didn't you guys just use Eigen matrices and normal float/double arrays instead of creating this Mat template?
I'm new to cpp, and I'm trying to recreate getting texture coords for the candide model in android as JNI, am I better of following python steps rather than this?

Hi,
We are using HoloFace to build a game. It used to work perfectly fine in the beginning but now on deploying we can't see the content at all. We tried it on two different Hololens and it doesnt work on both. (It worked day before yesterday and since yesterday it stopped working, we haven't changed anything at all). What do we do ?

Hi, I'm looking into updating the projects to the latest SDK and platform toolsets in order to produce an ARM build to run on HoloLens 2. Unfortunately the repo ships with pre-built x86 LAPACK libraries so I'm wondering if you can point me as specifically as possible to the source of the current libraries so I can look into making equivalent ARM builds!

Many thanks for sharing this great project!

Dear MarekKowalski,
When I search google with keyword 'face fitting' and I read a code and description about HoloFace of you. But I don't know exactly formula about perspective projection(vector to rotation matrix), how can I do find it? And when I want to switch orthographic projection, where I can find formula? Can you help me?
P/s: I am sorry when ask you that, but I am a newbie an this field.
Thank you for reading

Hi,
Great work, I tested the effect ,but miss some include file and model. Can you supplement some files?
Thank to your open-source.

Hi there
Is it genereally possible to use the depth camera with that app? I adapted the DLL accordingly and changed the manifest but somehow I dont have access.
Felix

Is it possible to track multiple face at the same time?

Per your implementation of your get3DGaussNewton function, one of the first things you create is std::vector Sdf. I am a little stumped as to why you are adding 6 more Mat objects on top of the size coming from blendshapes, as referenced from the line below:

std::vector<Mat<float>> Sdf(modes3.size() + 6, Mat<float>(curShape.height, curShape.width));

My initial guess is it could be tied to the number of blendshape objects you are using when I referenced the Mesh folder under Assets. Could this imply that we will need to increment this value should we choose to add more blendshapes? Could you please clarify?

Thank you.