ALPACA Scale issue

I have 2 models I am aligning with ALPACA extension. They are same in size. I have clicked “skip scaling” option and only went up to “Run subsampling, run rigid alignment” stage (without CPD non-rigid registration/deformation). However, I noticed that my source model (red) has shrank in size.

Why is this so? and how can I prevent this happening? I’d like to keep both at the same size at all costs.

Screen Shot 2021-06-28 at 1.58.37 AM

The funny thing is, this only happens some of the time. Other times, the shrinkage is slightly less, and at other times, they’re nearly the same size.

What do you mean by same size? Are they identical models?

Apart from the bit in the middle, yes they are identical prior to Alpaca alignment. They’ve been drilled slightly differentlt in the middle, but they are the same teeth. I’ve even used mark up functionality to detect the change in length (around 5% decrease of source model) of the red source model.

Can you share your data, along with the parameters you used in ALPACA?
@agporto

Yes here is the link: ALPACA - Google Drive

I’ve used the standard parameters with no changes, apart from skip scaling = checked. and only went up to “Run rigid alignment” stage.

Only the source model seems to be affected.

Please do have a go multiple times, as this happens only half of the time.
Also when I swap the source & target (target as the source, and vice versa), the issue seems to resolve …

Here’s another variation that I got, after increasing the point adjustment density to 1.4. @muratmaga

Screen Shot 2021-06-28 at 1.27.19 PM

After repeated attempt, I obtained this variation as well, with same parameters as before (this is what I’d like to achieve every time - no scaling done to source model)

Screen Shot 2021-06-28 at 1.31.08 PM

1 Like

Hi Sean. Thanks for reporting this. Some recent modifications to the code have broken the scaling behavior. I will correct the source code tomorrow (and ping you).
Thanks again

No problem Prof Arthur
I’m glad I could make a contribution to this awesome software.

1 Like