Problem with masquerade

[MaelstroM]

[Andrea Alessi]

Hi Maelstrom,

it looks like the problem is the fact that it is of a uniform color: Uniform colors are very difficult for Zephyr, as they have no feature to detect. If you need only the geometry, a quick fix is to dirty the surface in some way (with spray, or even flour for example). Otherwise if you need the texture, it's going to be more difficult. Can you share some pictures (or better, the whole dataset) so we can give you more specific tips?

[MaelstroM]

Hi Andrea, here you can find 4 samples of this dataset. Thank you.

For example DSC01287 and 88 are not recognized on the reconstruction process ( too much brightness i think, DSC01312and 14 are OK.

Not recognized ->



OK ->



Perhaps this work on a lightbox studio with a diffuse and cold light could help me to have a clean model to reconstruct and as you say : dirt the object or applying flour on it.

[Andrea Alessi]

A lightbox might actually be worse, as the lights would shine and create highlights on glossy, uniform surfaces.

I assume using flour is going to be a problem if you need the texture.

The first thing i would try to do is try to reshoot the images, but this time put a newspaper on the bottom, and try to always have the newspaper in your frame: this will help zephyr in the camera orientation fphase.

As for the dense cloud generation, you will still have some issues, because the surfaces is very neat and uniform, so i'll expect some noise. That said, i think you should be able to extract something. Try as i said above, and also try to minimize the highlights (no flash, no direct sunlight : make sure to shoot images with diffused lighting).

If you dirty the surface, you will get a better geometry (for example by spraying it with flour, or by projecting a pattern onto it).

[MaelstroM]

I think projectif a pattern on the subject could be the solution, a simple model like this one is easy to remap and texture.

A lightbox with filtered light and no direct light ( light will be filtered by a white filtering piece of textile ) should also work in this case, i won't have no highlights on the model and due to his uniform texture and surface, thee solution filtered light and pattern projection should help a lot.

I will make tests.

[Andrea Alessi]

if you want me to do some tests too, feel free to share your dataset with me when it's ready and i'll gladly give you my two cents

[Helmut]

I have encountered another problem with masquerade: I masked a set of 50 photos individually, but when I tried to save the masks I got the answer :"Failed to save masks. Please see log for details."
But I didn't find any log button.
This problem happened with silhouette masking as well as with lasso masking.
What to do?

[Andrea Alessi]

Hello Helmut,

you're right, it's a wrong error message. Unlike Zephyr, Masquerade does not have a log file.

It looks like masquerade had some permissions issues. My guess is for example you tried to load the photos from a SD card or similar, where you did not have permissions. Try copying the photos on your desktop (or saving the masks on a separate directory rather than the default one).

[Helmut]

[Helmut]

Just another newbie question: What are the pro's and con's of silhouette masking vs. lasso masking?

[Andrea Alessi]

Make sure that that dataset is writable by you. If masquerade can not access the file system, either the directory is read only, or the windows permissions are wrong (for example you may have copied it using another windows user) or external software may be interefering with it.

The silhouette masking is much faster and is exteremly useful especially for turntables or when you have high contrast background (for example, a human and a green screen). The lasso tool is useful to do quick-fixes or for other cases where the silhouetter system doesn't work that well.