Sketchfab Forum

How to merge the base/bottom of an 3D object on Agisoft?

(Panomedia) #1

Continuing the discussion from Best method for capturing all sides of an object?:

Add new "3D scanning and photogrammetry" group in forum
(Lapin) #2

You can see this lesson in three parts.

(Panomedia) #3

Thanks very much @lapin smile ..i went with method smile worked good for me smile

(Paulodisino) #4

Hey @panomedia, which method worked better for you? You kind of just created a link to send to gerpho's profile.

(Panomedia) #5

Hi @paulodisino gerphos's method was fine for me... that is taking pictures on a turn table and then puting the object on its side and have one picture that matches the same view as when it is standing up ... like the sample pictures gerpho has of a monkey ... i realized though that it doesnt always work as we wish.. and maybe need external 3d editing software ..

(Paulodisino) #6

Hi @panomedia thanks for the advise, is there a link to how @gerpho did it?
By the way, what are your pros and cons in your view in the subject: BobCapture vs PhotoScan?
Is any of them better when it comes to merge the two sides?
Did you try the @lapin tutorial?

(Gerpho) #7

Actually I mentionned this method, but never explained it in great detail.
When I want a small object to be visible from all sides, I put it on a turntable, shoot on a complete rotation (with the camera at a certain angle). Then place the object on a different side, and start again.
In theory two sets of pictures are enough, but I usually do more, to increase quality and avoid holes.
Here's an example of the shooting sequence for my Rhino Beetle : Notice that each circle intersects two others, so that connection between photos belonging to different circles is easier to perform (for the software of course).

Final result is here :

Rhinoceros beetle (Oryctes nasicornis) by Gerpho on Sketchfab

(Paulodisino) #8

Hello @gerpho, that is great!
Awesome idea and great detail on a piece so small like that.
Did you use any markers underneath it or just laid the beetle flat on a plain white turntable?
Thanks for sharing

(Gerpho) #9

Hi @paulodisino, thanks.
Interesting question (about the markers), the answer is absolutely no. The reson is that since I re-position the beetle in several different positions on the turntable, these marks would not be in the same place relative to the object from one circle of photo to the other, and this might prevent the software from connecting the different circles (if you see what I mean, difficult to explain). Therefore, I take great care to avoid all defects and marks on the turntable !
Actually I did try adding marks once and the initial reconstruction (AT phase of ContextCapture) was a real mess, with bits of the turntable surface floating in mid-air all over the place : the software was trying to make sense of these marks, without managing of course.

(Paulodisino) #10

@gerpho I see what you mean.
I saw this tutorial on merging chuncks and I just did a very nice statue full of details and all, but I left the marker below .. let's see if I can manage to work with what I have, otherwise it is going to be a mess like you said.
My decision to leave the markers on was because once I tried to scan a flint (only one row around 45o high) without markers and it was a big mess with all the cameras mispositioned all in one place. Later on I did the same scanning with the markers and it worked very well.
So basically following what you said I needed only to photograph different rows, separate in chuncks and then let the software figure it out for the merge part. Is that it?

(Gerpho) #11

Welllll, is't getting rather involved.
To simplify : ability to reconstruct depends on the software you use, some are more able than others on certain types of scenes. Then tolerance to moving markers too. And finaly, the object itself can make a big difference ; the mamal jaw I posted some time ago was difficult because it's rather thin, and I had difficulties connecting both sides. Other softwares can manage better maybe. I think context capture is rather weak if the scene doesn't cover the complete photo (as in the case of an object).

I think from then on, you have to build your own experience, there are too many variables involved for advice to be of any use.