Landmarks for shape completion (Tutorial 9)

20 views
Skip to first unread message

Maia R.

unread,
Mar 1, 2021, 2:54:45 PMMar 1
to scalismo
Hi,
In the Tutorial 9, 200 corresponding points were used to perform GP regression. How were those corresponding points obtained between the partial mesh (noselessLandmarks.json) and the model ( modelLandmarks.json) ? 

Were they obtained by means of non rigid ICP combined with Gaussian process regression as explained in the model fitting (Tutorial 11) ?

I am a bit confused : is the the posterior model here is the same as the last one obtained to establish the point to point correspondence ?

Thank you very much,
Best regards
Maia

Marcel Luethi

unread,
Mar 1, 2021, 11:42:34 PMMar 1
to Maia R., scalismo
Hi Maia

The crucial assumption here is that all the 200 landmark points are in correspondence. Otherwise GP Regression would not work.
We have indeed run the non-rigid ICP algorithm to establish correspondence between the given fragment and the model to obtain this correspondence.
Once correspondence was established, we simply picked 200 points at random.

The posterior model we compute in tutorial 9 is the same as we compute in the non-rigid ICP. For the non-rigid ICP we use, however, only the mean whereas in tutorial 9 the main point is that we obtain the full distribution (another GP).

Best regards,

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/86201e9a-54bc-420d-a666-2b036510340an%40googlegroups.com.

Maia R.

unread,
Mar 2, 2021, 1:12:07 PMMar 2
to scalismo
Thank you very much.
Best regards

Maia R.

unread,
Mar 3, 2021, 9:32:57 PMMar 3
to scalismo
Did you use known landmarks correspondence as regularization in order to establish the correspondence of those 200 landmarks ?
Thank you a lot,
Best regards,
Maia

Le lundi 1 mars 2021 à 23:42:34 UTC-5, marcel...@gmail.com a écrit :

Marcel Luethi

unread,
Mar 5, 2021, 2:13:05 AMMar 5
to Maia R., scalismo
Hi Maia,

I think we just run an ICP algorithm, where we sought correspondence starting from the partial target to the full shape. Once we had correspondence, we could generate the landmarks.
There are many different ways of doing that and, given that we did it 5 years ago, I cannot recall the details.

Best regards,
Marcel

V R

unread,
Mar 5, 2021, 4:28:41 AMMar 5
to Marcel Luethi, scalismo
Hi Marcel,
Thank you very much.
Best regards
Maia
Reply all
Reply to author
Forward
0 new messages