Advice on reflections which create holes

Hello!

I have the following problem:


You can see that from one side, the top metal is fine, but from another angle, it looks white.

This produces holes in my mesh:

This is problematic because the texture is missing (the correct texture is not used because RC likes to use the “white” texture instead), and of course I have to fix the mesh in another software.

I have played around with the alignment settings, but it didn’t help.

Do you have any suggestions for such a situation?

Thank you.

Hi tw2016

Just get more images looking at that particular area of the model … that is the only way…

Wishgranter wrote:

Hi tw2016

Just get more images looking on that particular area of the model … that is the only way…

I have updated my post to show you that I have more images of that area, but some are reflective, and some are not.
Can I attract RC to use some textures / areas?

tw2016,

That looks like it captured really well actually - what does the generated model look like? While it looks like some holes in your point cloud, is it translating all the way to model output?

I don’t see the reflective areas you are talking about but if you are getting glare and/or light is reflecting off of something like glass, you can use a circular polarizer to tweak the image taken - but I just don’t see how that scene is causing reflection errors. Stone generally captures really well.

Wishgranter is right, but if you still can’t get it to capture, that’s one of the easiest things to correct in post - a quick rectangular mesh box and cleanup of RC’s geometry and re-import into RC for texturing has a really good chance of making that model sing.

Castlenock,

the mesh unfortunately has these holes, and I’m not really familiar with fixes such large holes in ZBrush. The YouTube tutorials didn’t help much for these large areas. When I dynamesh, the other parts of the model become just too smoth, and I loose details.
I also looked around if any software could fix such hard surface geometry automatically, but without success.
You wouldn’t be in for making a YouTube video with your favorite mesh-fixing software, would you?

what wishgranter said, and maybe you could try a polarising filter with your camera to remove the white look,
the next thing what i would try is masking the object or just parts.

Nedo wrote:

what wishgranter said, and maybe you could try a polarising filter with your camera to remove the white look,
the next thing what i would try is masking the object or just parts.

Do you mean I should put a black area over regions in images that reflect too much, in an external image editor?

tw2016 wrote:

Castlenock,

the mesh unfortunately has these holes, and I’m not really familiar with fixes such large holes in ZBrush. The YouTube tutorials didn’t help much for these large areas. When I dynamesh, the other parts of the model become just too smoth, and I loose details.
I also looked around if any software could fix such hard surface geometry automatically, but without success.
You wouldn’t be in for making a YouTube video with your favorite mesh-fixing software, would you?

Heh, that would be very definition of the blind leading the blind. That and I chose Modo as my program of choice, which is great if I want to scale or make a business of this stuff (I do, it’s sooooo rewarding), but is quite expensive. I started learning on Meshmixer but I only removed stuff, didn’t create too many things.

I will say this though - I approached 90% of all of this with zero knowledge… I think learning the next stage of reconstruction is totally worth it though - it’s ■■■■ hard and is super easy to give up in frustration, but if you break the barrier and it clicks it’s totally worth it. That and your models will start to shine. Good news is you don’t have to learn the whole capabilities of the next step, just parts of it to become really good at reconstruction (I’m far from there yet, but I feel like I can attain that at some point now). Pick a poison and if you haven’t gotten a free trial, try Lynda.com for some training in it. I can help quantify some things that I’m still learning but are really hard to grasp for beginners like UVs, unwrapping, re texturing, etc. <- that’s what I’ve found to be the hardest by far as these sort of things seem to be geared for 3D modelers in general that are at a high level and you’ll hit that a lot earlier in your RC workflow.

If you’re not already following it, the Valve tutorials for Destinations are a great, fantastic start - only about 1/5th of the tutorials there are really about the Destinations application workflow, almost everything else is photogrammetry tips and bopping back and forth in that second program in your workflow. It can be overwhelming, but a great resource. You’ll probably learn the hard way, like me, that missing even the smallest bit in the tutorials will hurt a lot, but that’s part of the process.

In a long and storied career, I’m not the type of person who enjoys my work, I just enjoy the outcome/result. This is one of the few things where I have loved every part of the process (including the end result). I’m sure some of that will wear off if I make a living off of it, but if you like it, it just gets better as you learn more.

Thanks.

tw2016 wrote:

Thanks.

No problems, but I kind of came off there as knowing more than you and it’s probably the other way around, so sorry about that.

I’m not familiar with zbrush, but in Modo (or the free Blender) I’d just make a low poly rectangle as a separate mesh, superimpose it over that piece of wood, rotate, scale and transform it over the captured wood there, and once it is dead on after a lot of tweaking, delete the RC generated geometry, re-project a UV, and re-texture in RC . It comes out a lot better than you would think, still has a ‘real world’ feel to it, and you’ll save a ton on poly count.

I totally appreciate your advices, they help me a lot.
I thought the same: Insert a hard surface object into the missing holes and then merge it.
But it’s still a lot of learning, and being a game dev in my spare time only, I think I’ll hire somebody else to fix everything up for me.
But that is actually where I’m a bit disappointed about RC.
The software seemed so advanced that I thought that I could easily fix everything within RC as soon as I purchased a license, without having to use external programs
I had to learn that RC should be used to produce nice meshes only and that cleaning up should be done in an external program.
If I’m wrong, anybody might gladly tell me otherwise.
I played around with the settings so much, but whenever I used anything but the default values, my model got worse.

Having a tool that would allow for redefining areas that are more or less quads would be such a huge timesafer for everyone interested in game dev.
For example, I would just have to define 8 points to fix up a hard surface quad. That would be so amazing.
There are many objects that consist of pure quads.

Hi tw2016
Yes, you are right - RC is mainly for reconstruction of data to 3D models. Take a look at Zbrush – it is the main tool for postprocessing of 3D scanned data. Even though the UI does not look very familiar at first, after few weeks you will see how powerful Zbrush is for this sort of work…

Yes, I know ZBrush, but most of the work that I have to do there is very basic, nothing special:
Deleting parts of the mesh, auto-po and baking of normal and occlusion map.

That is why I think it should be part of RC.
This would make people’s workflow much easier.

Some other softwares (which are not as good as RC but still good) have tried including these work steps in their software.
I think that is a very good idea.

tw2016 wrote:

Yes, I know ZBrush, but most of the work that I have to do there is very basic, nothing special:
Deleting parts of the mesh, auto-po and baking of normal and occlusion map.

That is why I think it should be part of RC.
This would make people’s workflow much easier.

Some other softwares (which are not as good as RC but still good) have tried including these work steps in their software.
I think that is a very good idea.

I understand where you are coming from - I had the exact same opinion until I started my ‘big’ project: that project has outside scans, whole rooms, small exhibit items, big exhibit items, etc. etc. I picked it as it kind of encompassed ‘what is the hardest things you can scan, and what can cover all sorts of scanning scenarios’.

I also had the exact same opinion of ‘I don’t have the time to learn this other stuff, I just want to focus on the scans’. I have changed my opinion on that, however: mainly because after so much work you really need the Swiss army knife of a program like Modo (or Zbrush). I think trying to reproduce that in the scanning software starts to half measure the dev’s times on getting the photogrammetry software top-notch. That and they could never serve their demographic - you and I may want a tool that does A, but their customer base needs tools that do A-Z, which all of those second programs are designed for.

That’s not to say that I wouldn’t love to see some of that stuff seep in, like Meshlab’s ‘fix-holes’ which seems to be a requisite step for any of the big photogrammetry hitters like RC or Agisoft. Ultimately however, I want RC/other-programs to improve on the photogrammetry bit first and adapt new technologies as they come available second (like project Tango, or other things that could could augment Photogrammetry, etc.).

That’s just an opinion though - I think publishing RC on Steam somewhat switched their customer base to newbies like ourselves; I wouldn’t be doing or know any of this if it were not for Valve’s tutorials for Destinations and I honestly believe that in a few years the demand for scanning areas is going to spike for non-tech clients.

hello, have you already tested RECONSTRUCTION / Tools / Close Holes?