Mesh not correct, surface is extended

Hi,

I’m trying to reconstruct an interior, so I take a lot of photo from inside the room.
When I export the 3D mesh, I notice that there is an external enveloppe for the mesh thta has a form of a big box. I looks like the software try to make a closed mesh (or a topologialy correct mesh). I just want to create the surface, but dont want the extra part of the mesh.
Is there any option for that? (This is standard in other photogrammetry softwares but not in RC.)

I hope you will understood what I mean and can help me.

Thanks
Ben

Mar wrote:

Hi,

I’m trying to reconstruct an interior, so I take a lot of photo from inside the room.
When I export the 3D mesh, I notice that there is an external enveloppe for the mesh thta has a form of a big box. I looks like the software try to make a closed mesh (or a topologialy correct mesh). I just want to create the surface, but dont want the extra part of the mesh.
Is there any option for that? (This is standard in other photogrammetry softwares but not in RC.)

I hope you will understood what I mean and can help me.

Thanks
Ben

Hello Ben,
go to RECONSTRUCTION tab and look for Filtering tool. With this tool you can easily remove unwanted traingles.
There is also topic in the application help, how to use it :wink:

Hi Kruzma, Thanks for your answer.

I tried by I was not able to do it (after reading the help). When I follow the instructions for creation a reconstruction box, nothing happen, it only make my model pan. In addition, as there is no display of the polygon so how can they be efficiently filtered ?

In general I find the software very confusing and not intuitive. The algo and the results are really great, probably the state of the art, but the interface is not logical at all (unless for the ones who made it). I Wish that they hire a good UX guy to redo it. (But this is not the topic of this post, I should make a long wishlist in another). The fact that we have to go often to the help menu indicates that it’s still not intuitive. (I nearly never looked for help in 3dsmax or photoscan)

I will give again a try.
Thanks
Ben

Hi Mar
it is much more about understanding how it works and you probably find it complicated but it is just the opposite…
To solve this, just set a bit SMALLER BOUNDING BOX (BBOX) than in the actual reconstruction, then use FILTERING and set SELECT INSIDE or OUTSIDE of the BBOX and use APPLY… try it, it works and it is trivial…

Hi Wishgranter,

Thanks for your answer.

For the Box, I tried to do it but it’s not working in this release for me :
When I click “Set Region Manualy” and left click on the 3D view it’s only paning…
Strange thing : I can edit the reconstruction region when I select Ortho Projection. I’m still very confused.
When I right click on the reconstruction region, one on ten time I can display the gizmos to edit it. Same thing, I can’t get the logic.
I wish there where contextual menu in the 3D view to “edit reconstruction region / clear region /…”

I read that I’m not the only one that is confused by the organisation and the logic, ie :
Why having same functionnalities in differents places ?
Why having major topic writen on top :workflow/alignement/reconstruction… and the sub topic written on the bottom (1.add Imagery/import metadata/2. process.) ?
Why my active component is not highligted?
Why can’t I rank camera poses by names/ by date/ by number of tie points (It’s so frustrating to look for a specific image in the list) ?
Why can’t I display camera poses list as a row of thumbnails (I so usefull)?
Why can’t I move from one shot to another by pressing up/down arrows?
Why can’t we select and display 2 components at the ame time?
Why can’t I higlight the picture in the list by clicking the camera pose in the 3D view?
How to get back the Quick info panel?
Why there are some element with on mouse over info and some not ?
Why undo is not undoing the last action?

I could go on…

When it’s about usage, users are not wrong. I’m currently going back to photoscan even if it’s take 100 more time and the quality is not as good as yours because I can’t succed in doing simple reconstructions. I’have been doing 3D for over 30 years, testing a lot of softwares and yours is (for now) not intuitive (but one of the most promising one). You should hire a super UX designer and you will win the market.

I hope you understand that I’m writing that not to freely criticise but instead to increase product logic/quality/usage and productivity.

In the wishlist : do you plan to implement ICP algoritm to avoid manual component alignement? Do you plan to add video as source?
Do you know when the 3D mesh display will be released?

Hi Mar.

I read that I’m not the only one that is confused by the organisation and the logic, ie :
Why having same functionnalities in differents places ?

  • advanced control.

Why having major topic writen on top :workflow/alignement/reconstruction… and the sub topic written on the bottom (1.add Imagery/import metadata/2. process.) ?

different workflow, like ribbons in 3DMax or rooms in 3DCoat.

Why my active component is not highligted?

When you choose component or image it become Highlighted.

Why can’t I rank camera poses by names/ by date/ by number of tie points (It’s so frustrating to look for a specific image in the list) ?

  • Camera in RealityCapture in two forms. Aligned camera - useful data. And analigned.
    It can be some other grouping factor like EXIF data (different camera) and that’s all.
    Just don’t mix in one project different datasets, closecaptured turntable apple and castle scan with aeral/DSLR/lidar.

Why can’t I display camera poses list as a row of thumbnails (I so usefull)?

  • Switch left window to 2D view

Why can’t I move from one shot to another by pressing up/down arrows?

  • Use left/right arrow in 2D view.

Why can’t we select and display 2 components at the ame time?

Components can have different scale, so this is useless in 99% cases.
Treat many components as error of person who was acquired images for reconstruction.

Why can’t I higlight the picture in the list by clicking the camera pose in the 3D view?

  • Can, please check manual.

How to get back the Quick info panel?

  • Click on any object in project.

Why there are some element with on mouse over info and some not ?

  • Some objects can have additional useful information, that can used for inspection.

Why undo is not undoing the last action?

  • RC do not delete nothing. All “destructive” actions just create new object that you can delete. Nondestructive - you can fix as much as you want.

I could go on…

When it’s about usage, users are not wrong. I’m currently going back to photoscan even if it’s take 100 more time and the quality is not as good as yours because I can’t succed in doing simple reconstructions. I’have been doing 3D for over 30 years, testing a lot of softwares and yours is (for now) not intuitive (but one of the most promising one). You should hire a super UX designer and you will win the market.
I hope you understand that I’m writing that not to freely criticise but instead to increase product logic/quality/usage and productivity.

All RealityCapture UI/UX have solid logic. Based on in-house R&D and real life photogrammetry scans of objects from simple closecapture to castles and aeral.

Let me say figurative, all time you worked at road-roller. It strong and solid. RealityCapture is airplane. So if you want a fly you need learn something more than how to drive roller. And after this you’ll fly :wink:
RealityCapture UI/UX allow you control all required settings for full control of all aspects of photogrammetry process. And if you want use all this power you will need learn not only this software but better to learn all photogrammetry basics, from acquiring images (that is 90% of success) till some internal photogrammetry mechanics.

In the wishlist : do you plan to implement ICP algoritm to avoid manual component alignement?

ICP required good (about 90% overlap) between cloud, so if you have such good overlap in your source images you don’t need ICP because RealityCapture will create “solid” mesh from start.

Best regards.

Mar.

For import images to RealityCapture you can use opensource ffmpeg.
It will give much more control even if Capturing Reality implement video import in RC (if not yet)

Check this examples:
https://trac.ffmpeg.org/wiki/Create%20a … he%20video

if you have MJPEG as video source you can use this settings for extract frames as JPG without recompression:
http://stackoverflow.com/questions/1022 … ith-ffmpeg