First sets of photos will only align 3 photos.

[fodder1]

Hi Everyone,

so complete noob here when i t comes to this. but you have to start someplace.

So i downloaded and printed the $30 3d scanner from thingy-verse. Basically it is a turn table that takes 50 photos per revolution.

The first and second try were aligning the part and getting to know the interface etc.
But two other trys have led to the the same result, only 3 of the fifty photos (free version) are being aligned.

I did a bit of reading that the surface of the turntable (its LIME GREEN) might have been the problem. to fix this i cut a bit of white paper the same size as the turn table.
the other suggestion was lighting. I don't have light tent, so positioned a single LED light directly above the turntable so there is minimal shadows.

I have included a quick photo of the capture rig, please let me know what i am doing wrong.

[Andrea Alessi]

Hello fodder1,

it's hard to tell without looking at your photos. However from what you say, i think you simply need to mask your images.

If you do not mask (use 3DF Masquerade - it should take just a couple of minutes using the automated strokes + turntable mode) your images, then zephyr is confused because the object move and the scene is no longer coherent. Do you want to reconstruct the object, or the table/wall ? Zephyr has no way of knowing it unless you tell it what to ignore.

If you'd like to share the images, i'll happily have a look.

[fodder1]

Ok, that makes sense.
I will see if i can run the photos through 3DF Masquerade and see how they go.

thanks for responding. i will report back.

[fodder1]

ok,

i ran 3df masquerade and masked all 50 photos and the started a new project and loaded the new photos.
I selected the "mask" option and the 3df masquerade files for all 50 were loaded.

but i still only got 18 files, after running.

I tried to up load all of the files, then 10 and then 1 with no luck. the server just sat there spinning.

i believe that i will have to work on the lighting. Probably by a light tent and maybe some led strip.
but that does not explain why it seems to only be choosing the middle of the group of photos.

Also i could use a Nikon D70S that i could use instead of the iPhone. Yes it would probably be better but i have to start some place and this turn table and the iPhone camera is fairly good.

So short of the shadows which could be fixed with a bit more light, what am i doing wrong?

[Andrea Alessi]

Hi fodder,

it's hard to tell without looking the images. Can you share a few?

[fodder1]

i am trying but the server wont let me upload them.
It is saying they are too bit and need to be 600x600

[Andrea Alessi]

Can you upload them to somewhere public like imgur or dropbox? If not, i'll send you my email address privately so you can send a few there.

[fodder1]

ok, a couple resized to about 15% or 450x600 from 3024x4032

[fodder1]

we can do dropbox for the full images if we need too.
its late here in Australia, so gotta head off to bed now.

[Andrea Alessi]

Ah i see. Well, it's hard to tell with the resized images, but it looks like a challenging subject. Photogrammetry works well with textured subjects, plain, uniform surfaces are going to be hard. You should try dirtying it or using a projected pattern (which will be tricky using a turntable setup, see here https://www.youtube.com/watch?v=E06kgYBftak at 4:20 for an example).

Anyways yes, when you can please share the full dataset, i'll have a look at it

[fodder1]

i will put them in a a dropbox account, contact me privately, and i will give you access.

There will be a lot more bits in the future i am sure, so learning now is good. I really only need the basic shape to stick it into a CAD program.
from there i can massage it into shape and print it.

[Andrea Alessi]

Hi fodder1,

I sent you a PM. Please send me the dropbox link over PM, thank you!

[Andrea Alessi]

Hi fodder1,

so i had a look at your dataset.

A few suggestions:

• when you need to reconstruct uniform colored objects, you should consider dirtying the subject (e.g. with powder) or with a projector.
• if this is not possible, you should put something textured (e.g. a newspaper) under the object to help the orientation phase rather than a white sheet
• try to use a diffused light - unfortunately your subject is already challenging (uniform color, no texture, small) and direct light will create highlights
• consider, if possible, to use a better camera. Unfortunately an iPhone is not the best tool for close range photogrammetry for this type of subjects. It will work for easier textured subjects, but don't expect too much for this type of objects

So while i could get to orient a bit more images, it is not worth investing more time. Please try to re-acquire your dataset and try again!

[fodder1]

sorry for the time delay,life getting in the way of fun.
I have access to an old Nikon D70 that might be able to do a better job on the photo side of things.

the rest of the points are easy to test out, like changing the paper to news paper and maybe a little cooking oil spray on the part and a bit of flower powder covering the object.

through that in a light tent with some diffused light and we will see how we go.

thanks for your time!