Quality Images of a Toddler?

[T. Dious]

[CG-Guy]

Living subjects are impossible with one camera because if they move or anything changes from pic fo pic it wont reconstruct properly

high end companies have a stage with cameras surrounding the subject to take a snap at the same time.

[Andrea Alessi]

Hi T.Dious!

thank you for the kind words!

As CG-Guy pointed out, living subjects are difficult with one camera only as they tend to move. A good idea is to use a video rather than a sequence of images, so you can be quicker to capture the subject and mimize movements (here is a great tutorial by Nick Lievendag https://3dscanexpert.com/free-3d-sca...eo-smartphone/ )

Using a video has the additional advantage that if you're scanning a sleeping toddler or pet, they won't be woken up by the CLICK of the camera.

On a side note, 3DF Zephyr has an experimental support for synced camera (DSLR canon only - 0 for free, up to 2 for Lite, more for pro/aerial https://www.3dflow.net/remote-dslr-s...ad-3df-zephyr/ )

[AvionAvionics]

Maybe while the child is sleeping? I agree with CG-Guy that a camera rig is the best solution but if you're poor like myself then that may not be an option (not to be that guy but I prefer having a house over 50 cameras lol). Video may be easier for you as well instead of an actual camera. Even with my Samsung Galaxy Note 9 phone, I have managed decent renders in video mode

[T. Dious]

Thanks for the replies.

With the success we've had so far, we are about to begin playing "games" of "statue" with the kid in hopes of eventually getting something useable.

In addition to the video suggestions, we have access to three identical cell phones who's cameras could be set identically and carried around the kid simultaneously taking either stills or video.

What is the likelihood of getting anything usable out of a good number of cheap ($5) camera boards from AliExpress such as below? Obviously, you would need a processor and code to control it...

https://www.aliexpress.com/item/FPV-2-8mm-LENs-1-3-Pixelplus-PC1099-CMOS-image-sensor-camera-PCB-board-module-for/32861903466.html?spm=2114.search0104.3.52.790f2d88nuIFPO&ws_ab_test=searchweb0_0,searchweb201602_5_10065_10068_10130_10890_10547_319_10546_317_10548_10545_10696_453_10084_454_10083_10618_431_10307_537_536_10059_10884_10887_100031_321_322_10103,searchweb201603_52,ppcSwitch_0&algo_expid=1bfbaff2-9708-4a4c-a0ea-bbea772b741e-7&algo_pvid=1bfbaff2-9708-4a4c-a0ea-bbea772b741e&transAbTest=ae803_3

or this basic type of thing.

[T. Dious]

Fixed link above

[cam3d]

I can't speak for those specific camera sensors, but I managed to build a head scanner using 5mpx Rpi cams - It's not easy, it's not cheap and I wouldn't personally do it again.

You'll get much much better results from a DSLR with video functionality, rather than a cellphone in my opinion. Bigger the sensor the better (generally speaking )

[T. Dious]

Thanks for the DSLR/ Video suggestion. I knew the Canon T6 would shoot video, but for some reason it didn't occur to me to try it.

I need to do some more testing, but I seem to be getting better results from cell phone video (Huawei Mate SE) than my Canon T6 still images. I must be doing something wrong. Lol

Fun stuff.

[cam3d]

I think it's probably the speed at which you're acquiring the data with the cell phone (probably faster, so there's less movement) - Feel free to share your results here! - What kind of camera settings are you using with the T6 photos?

[T. Dious]

I expressed myself poorly above. I was typing on my phone, and in an effort to be brief, I failed to communicate. :-)

We haven't yet tried any photos of the kid, as I want to have a high probability of success from the beginning. Little ones can tire of such things quickly.

What I meant to say was that I seem to be having better success with the phone video than with T6 stills to capture an inanimate object, but it's hardly been scientific so far.

Different objects, backdrop, time of day/ lighting, etc. don't make for good comparisons, so more consistent tests are in order.

Most of what I've been trying to capture is in my too dim (though relatively well lit) kitchen, which I think is a tough environment for this, but it's what I have readily available.

Models have included a red leather purse, a black Haflinger brand wool clog, and most successfully, a full color 3d medical model of a human brain. The clog may have too little texture, the purse too little texture and a shiny surface. But, since I've been pretty random with things, I really don't know what's causing what.

WRT the T6 settings, I have been running ISO100, and Aperture 22, whatever shutter speed gets me a centered exposure meter, maintaining a constant focal length, allowing the camera to auto-focus, while moving the tripod around a rectangular table, shooting with a remote shutter trigger. I think this seems reasonable from what I've read?

Once I get some better data, I may post asking for specific input on process improvements.

Feel free to comment on the above, of course. ;-)

Thanks for everyone's input.

[cam3d]

A few little things:

ISO 100 isnt the be all and end all of quality - I frequently get up to 3200iso and still get nice results (albeit with a full frame camera) You can comfortably get up to 400iso with the T6 without much noise at all. Even with a turntable, pumping up the ISO a bit to speed up the capture can definitely be a bonus.

Id argue that f/22 is crippling you a bit, and that somewhere between f/11 and f/16 would be better (diffraction is real, but minor - If your whole subject is in focus at f/11, you will see noticeably sharper photos than at f/22) - Shutter speed could be an issue if theres any vibration (but youre using a remote so vibration is much less of an issue) - I recommend 1/160 and up for handheld.

I hate scanning people handheld. Its the uncontrollable human element which gets in the way

[T. Dious]

Interesting. I will try these suggestions as soon as I get a chance.

I will confess to being almost completely inexperienced when it comes to photography. I was aware of the exposure triangle, but never used it before taking this on.

This is my wife's camera and it may never have been on the manual setting before I tried photogrammetry.

Again, I appreciate your input.

[Victor Ramamoorthy]

A few years ago, I constructed a 3D scanner with 40 Sony cameras with a lot of wiring and control hardware. I would not suggest that approach today. What is easy is to build a real time 3D scanner with Raspberry Pis and RPi cameras. There are various designs on the net that you can consider. The other approach is to buy structure sensor from www.structure.io/structure-sensor for ipads or structure core for rest of the platforms. I don't have any personal experience with this, but looks promising for the cost ( approx $ 500 to $600).

[T. Dious]

A few years ago, I constructed a 3D scanner with 40 Sony cameras with a lot of wiring and control hardware. I would not suggest that approach today.

Thanks for your thoughts.

The Rpi stuff looks interesting, but I would have to scale up a turntable for my intended subjects. That's a maybe. ?

Do you have a write-up anywhere on the 40 camera approach you took? I understand that you wouldn't recommend it, but I'm curious about what you did.

Hopefully, I will get a chance to try out some of the "camera only" suggestions given in this thread soon. Weather and life get in the way of fun.

Again, thanks to all of you.