Noseless Vetter landmark correpondence

[Maia R.]

Dear Marcel and dear community,

I am working on shape completion and I was wondering which method did you use to establish correspondence between the Vetter noseless landmarks (modelLandmarks.json and noselessLandmarks.json) in the Scalismo dataset.

I tried to get the same correspondence by means of non-rigid ICP with the analytic posterior. Lastly, I also tried to use the ICP-proposal of the github repo with high and low resolution model but I really can't have the same result with either the two approaches.

Could you give some advice/trick to do that please ?

Thank you very much.

Best regards.

Maia

[Marcel Luethi]

Hi Maia

The difference between our results and the one you are achieving could be that we used the full basel face model, whereas the data that we provide with the course is only composed of 10 faces. Consequently, the reconstruction won't be accurate.

You can download the Basel Face Model here:

https://faces.dmi.unibas.ch/bfm/

It is free for research if you use it only for research purposes.

Best,

Marcel

--
You received this message because you are subscribed to the Google Groups "scalismo" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scalismo+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/scalismo/b8fc43f6-b529-4b5c-bb0e-61e84ad14d50n%40googlegroups.com.

[V R]

Hi Marcel,

Thank you for your answer. 

What I mean is that I used the bfm.h5 model with the two approaches (non-rigid icp and the cp-proposal)  but I don't get the same point correspondence as yours.

What are the good techniques to achieve such a good correspondence (multiresolution, hyperparameter tuning ...) since the model is a very good one ? 

Thank you,

Best regards

Maia

[Marcel Luethi]

Hi Maia

Unfortunately, I cannot really tell you what we do differently than you. If I remember correctly, it is an absolutely standard ICP algorithm, where we sample points on the target mesh to avoid finding points without correspondences.

Best regards,

Marcel

[V R]

Thank you for your answer. 

I will do more tests then.

Best regards,

Maia