Marek

Hi, Well in order to do that you would have to segment out her hair, if it only has to work on this single image you can try doing it...

Hi, Performance is directly dependent on facial landmark detection. The landmark detector used here (the one from dlib) requires face detection in every frame which really slows things down. If...

Hi, The mean3DShape should be the same as the Candide neutral face as far as I remember. In the issue below I am posting a link to the piece of...

Hi, Sorry for not answering for a while. You can just use a weighted sum of the mask and the image if you want transparency., I am however not sure...

Hi, The content of candide.npz is basically a processed version of the original candide model you posted in your link. - mean3DShape: this corresponds to the neutral state of the...

Hi, Congratulations on converting the app to C++! There might be several reasons for this behaviour: - The 3D model vertices that are moved by the shape unit you added...

@vudangthinh Did you add any new shape units that would close the eyelids? If not, that might be the issue., I'm not sure if the original model had any. If...

@xiexiexxs Hi, I convert the shape and action units into blendshapes, which are S_i in the equation that you linked. As for the fitting function, yes that is the one...

Hi, I do not have any other implementations. I am however pretty sure that none of the model fitting code actually requires windows-only libraries. Marek

Hi, Below I am pasting a link to a very "dirty" script that does the conversion to blendshapes, so you can check it out yourself. https://www.dropbox.com/s/jbxqvm4o0qwv4fb/candide.py?dl=0 Best regards Marek