How to make scan model and texture densely registered to the 3D model

patrikhuber / eos

A lightweight 3D Morphable Face Model library in modern C++

Apache License 2.0

1.9k stars 597 forks source link

How to make scan model and texture densely registered to the 3D model #158

Closed zaykl closed 6 years ago

zaykl commented 6 years ago

Hi Huber,

First, thx your work. Recently I have read your paper <A Multiresolution 3D Morphable Face Model and Fitting Framework>. And I think the section 2.2 3D scan data is useful for us build our new model. I have found lsfm framework. but this framework don't support texture registered to the model.

Can you please let know some more detail about this work.

patrikhuber commented 6 years ago

Thanks!

I'm not really sure what your question is. If you're looking to densely register 3D face scans, eos doesn't currently do that. Please feel free to re-open if you have a more specific question or issue.

zaykl commented 6 years ago

Can you share some opensource project or paper working on densely register 3D face scans? Thx.

zaykl commented 6 years ago

I read your paper. While fitting 3DMM with photos, you train the SFM with scan faces dataset. So I would like to know how to train the new SFM by using the dataset I have.

patrikhuber commented 6 years ago

If you need to register 3D data/scans (i.e. bring in dense correspondence), you can try scalismo from Basel, the N-ICP in menpo, and I think the lab of Wuhrer et al. have some code online as well.
If your scans are already in dense correspondence, you can use the pca functions in eos (also available from the python bindings).