How can I improve this bronze statue photo scan?

(Paulodisino) #1

I am trying to improve this bronze sculpture's photo scan:

Right now I have done 7 chunks:
Top 57 photos
Mid-top 65 photos
Middle 49
Mid-bottom 36
"Mid-mid-bottom" 51
Plus 44 photos in 45o with it facing down and 43 photos in 45o facing up for the bottom part to be merged later.

Since I can not really put it on a stand because it is heavy and could fall, I had not choice but to shoot it in different angles and skip more photos from lower ground.

I know it is not an easy piece to do, but I figured if I can conclude this one, other ones won't be so complex.

So, as you can see by the photos compared to the scan, there seem to be a great deal of messed up poligons. I set up light so that it can also help show up the imperfections.

There is a bit of a problem at the bottom part since it seems that the exposure changed a bit and the merging of the chunks (even though they were aligned) really show the poligons of where I cut it.

I did some deleting at the dense clouds stage to get rid of some messed up data, but still some white background still spilled on the model.

I didn't use any mask.

The Dense Clouds processs took 4 hours. I got a Macbook Pro, El Capitan, I7, 16 GB, SSD 256, AMD Radeon HD 6770M 1024MB.

Please feel free to comment and question. Any advise, critique or suggestion is welcomed.

Thank you so much!

(Abby Crawford) #2

Cool piece, @paulodisino! What software are you using to process your model?

Your pictures look great, so my recommendation would be to mask them. That will eliminate the need for so many chunks and should fix some of the problems that you're having with polygons. Where shiny and/or complex objects are concerned, I have found masks to be invaluable.

(Matthewbrennan) #3

I assume you're using Photoscan, since you mention chunks. I would suggest you merge the aligned dense clouds into a new chunk, before making the models. Then trim your combined dense clouds, and create the model from that: this will eliminate the "shells" effect you see at the bottom and sides of your model.

With almost any photomodeling software, it's unusual to get a "perfect" result straight from the software - you almost always have to clean it up or edit it in another 3D modeling software. For something straight out of Photoscan, your model looks pretty good.

(Paulodisino) #4

Hey @matthewbrennan thank you so much for the advise.
I am gonna try your method.
Yes, I forgot to mention it is PhotoScan. So which software do you think I should use to continue and finish up the modeling?
I have Meshmixer because I wanted to try and print the 3D image of that statue, but I am new to it also, so I played a bit with it and found it was actually moving the texture a lot too.
I don't know if it could work, but maybe if I did the head, torso, legs in separated 3 chunks each and then merging them into a final model, then the level of details would probably overcome the bugs.. of course it is a lot of work.
If I did go through all that would I be able to do a nice 3D printing taking in account the complex amount of carved trails etc?
How are 3D printing coming out of photoscanning?

(Matthewbrennan) #5

You can definitely get good, detailed prints using a model made with photo-modeling.

Of course, the detail/resolution of the final model depends a variety of factors: quality of the images, resolution of the images, quality settings during processing, etc...

If you are just worried about printing it, forget about the bottom of the statue, just chop off the bottom and use a "close holes" command in Photoscan or other 3d software to make the model "watertight". Usually I make 2-3 versions of a model: 1 for web display (low resolution, retopologized), 1 at high-resolution (no decimation), and 1 for printing (medium resolution, unnecessary parts trimmed off, bottom flattened, etc..)

The key for 3D printing is that you eliminate any "funny" geometry - like the coplanar polygons / shells that you have in your initial model. Things like that can cause problems either at the slicing stage or in the actual print.

MeshMixer is a good starting point.

(Paulodisino) #6

Great! I got get to study these different versions and try to do or or two prints.
I had a goal to make the printed statue as close to reality as possible, but I guess that even with the tech now it is quite an impossible task.. or is it not? I mean.. the printer must be very well calibrated, there is always the necessary trimming and the size of the print might not be big enough.

(Matthewbrennan) #7

How large is the statue? Most consumer 3d printers can't print larger than 20 to 30cm in height. It is also harder to print in full color, unless you order a print from a online commercial printer. If you are using a printing service, they may be able to optimize the model for you (at an extra cost).

(Paulodisino) #9

Hello @jacobvr, thanks so much for the advises!

Right now the light set up was not the ideal. I live in an apartament with a living room that has a very big and "uniform" source of light, the balcony.
So I use it during the cloudless and bright sunny days during the afternoon.
This is my 2nd model so I am basically testing and now I am thinking of a more professional guear.
I was thinking in a 3 light set up, but I am trying to figure out what would be the better:

  • Isosceles triangle and subject in the center - 2 light through umbrellas behind the subject and a softbox in front of it. But I am not sure if I shine the light through a white cloth or if I go to the next model which I think would be the one you mentioned.

  • Isosceles triangle and the subject on the edge - 2 light through umbrellas side-by-side to the subject and a softbox in front. This would open space for the vinyl as you suggested.

Any ideas?

Add new "3D scanning and photogrammetry" group in forum