Adding photos of the underside in an open project

[kgf]

First of all,thank you for your time reading this.

Im a newbie...so im pretty sure this question has been answered 100 times already but for the life of me i cant find the answer to it...
now that thats outta the way...

im trying to create a model of a small rock.I have added 280 pics taken from a video.Now,the video was shot from above and around the object,and the underside was shot by turning the object bottom up.
I can see the program detects all camera angles as being from above and around,but not bellow the objects underside.

Hence my problem.The program uses the added bottom side pics but just adds them to the existing mesh without knowing they are supposed to go on the underside.

How do you get around this? cause the only way i can think of is to shoot it from below whille its sitting on a glass surface or something.

bare in mind i havent had any problems with the backgrounds.I use a turntable with a sheet of text stuck to it and it seems to be working fine.Ihavent had too much trouble removing unwanted points from the mesh.Seems to me ive done a pretty good job removing them manualy,at least on this simple subject,but the underside is really getting me stuck.

I understand this is a stupid question....so...go easy on me please...

again,thanks for your time.

[kgf]

Just as an update,my problem got magicaly solved just by changing the background on the turntable from something with features to completely white.This goes against every photogrammetry video i have seen so far i think,but for some reason i dont understand it worked fine the first time.the program sees the subject as elevated or something and puts cameras on the underside of the subject too.

i will try it again,but i was very pleasantly surprised cause i wouldnt have slept last night if i didnt get it to work

[Andrea Alessi]

Hi kgf,

everyone's welcome here, don't worry about your questions - we're happy to help!

You have to keep in mind that zephyr works by matching common points in the images - if the subject moves incoherently vs the background, there's your issue - the software can't know if you want to reconstruct the rock, or the background. The subject must not change overtime.

You've already found a possible workflow on your own: using a featurless backdrop so that zephyr doesn't pick up features on it and just reconstructs whatever is moving. Still, those pixels are still processed and might bleed in especially in the texture phase. So while it is a valid workflow, make sure your backdrop is really featureless.

The best way to process this type of "turntable" datasets, is to use 3DF Masquerade to mask out the background. This is a semi-manual operation (for now - great new tools are coming to aid in this specific part of the workflow with the upcoming 8.0 release next autumn ) but won't take long especially if you have a featurless background - you can use the red/blue strokes for a quick way to remove backgrounds on all your images. This has the added benefit of effectively ignoring parts of the images, so the processing will also be faster.

In some cases, you actually might want the background, but that's a different topic.

You can learn more about masking here:

https://www.3dflow.net/technology/documents/3df-zephyr-tutorials/tutorial-use-masking-tool/

https://www.3dflow.net/technology/documents/3df-zephyr-tutorials/tutorial-12-masking-for-turntable/

[kgf]

Andrea,thank you very much for your reply! Yeah i was trying to avoid getting to learn madquarade too cause i was a bit overwhelmed with the learning curve last night( tho it wasnt really that bad,just everything being completely new i guess) but yeah masking was gonna be my last resort.Thank god it didnt have to,so i got some sleep😄

Im gonna have to do it anyway tho just to learn the process.
Again,thank you very much for the reply🙂