34 out of 76 "cameras" used

[Andrea Alessi]

unfortunately some dataset will always work better out of the box in one software and other datasets will always work better in other software out of the box

All datasets are different (as all software are different) and ultimately, require the user to understand how to proceed depending on the outcome and desired result.

I still would recommend masking for better results, regardless of the software you're using.

[studio3d]

You're right! I'm on a 23" screen, I have to wonder why it opens less than full. I would never have known that stuff was hiding under there!

[AvionAvionics]

You need to widen your center panel.

[studio3d]

I looked at the tutorial on Masquerade, and I don't see the blue and red brushes in the app

[studio3d]

I just tried the same dataset in Agisoft Photoscan and got a very good model (needs some fixup) with zero masking. That's painless!

[AvionAvionics]

Someone may have a better answer than me but I did the same as their tutorial.

https://youtu.be/dGRw8LbXknU

Where I used the red brush to get the main object and the blue to ignore the background (turned on the turntable and always applied mask to next image) (Now i wasn't 100% perfect on my mask) but this route took me about 1-2 hours to do 135 images. It isnt automatic but it gave me results that not even deep scan could achieve by itself.

[studio3d]

Maybe I'm not doing it right, but masking is not painless! I tried to pick by color, but I had to add areas and subtract areas. Thought I could apply to all the images, but that sure didn't work! If I have to go in and mask all 57 images individually that would be very painful! I did a similar dataset on same background through Autodesk Recap and it AUTOMATICALLY ignored the background. I don't like paying subscription or "cloud points", so I was hoping Zephyr would be the answer...

[AvionAvionics]

studio3d give it a try, I was hoping to avoid masking at first but the time i spent masking has made up for the time needed for deeper or additional scans. Not only that the masking is quite painless

[studio3d]

OK, thanks... I will try the masking (though since the background has no texture whatsoever, I don't understand why masking is needed.)

[cam3d]

Hey Studio3d -

You're going to need to mask your images! - The image quality is much better now, so 3DF Zephyr has picked up on the background and aligned the cameras to that, instead of the actual subject. This is due to the DOF reaching further. You can quickly mask in Masquerade using the mask by colour feature

[studio3d]

I reshot the Lennon statue w iso200 1/2 sec f/20... used the jpgs which looked fine, and all 57 of 57 images were used... BUT the Sparse Point Cloud looked very odd, and when I tried to make a Dense Cloud the program hung up about 90% through. Here are the new photos... thanks. https://www.dropbox.com/sh/zmgxglobh...HdtsnYILa?dl=0

[cam3d]

studio3d

Andrea covered most of the important points - Just wanted to add, if you end up using your zoom lens, make sure you tape it so it stays at a single focal length, mixing focal lengths can complicate things a little and confuse the alignment. 60mm lens is perfectly fine, it depends more on the size of the object. For small things, a higher focal length is almost always needed (I use 100mm on insects and other tiny things)

You'll definitely see a noticeable improvement with a low iso, and if you're going for ultimate sharpness, you can use a DOF calculator https://www.cambridgeincolour.com/tu...calculator.htm to calculate optimal f/stop - Here is some interesting reading if you want to go deep: https://www.cambridgeincolour.com/tu...hotography.htm

[studio3d]

Thanks for your helpful hints, Andrea! My sharpest lens is the 60mm, but the suggestion on .pdf is to use 18-35... I have a zoom lens that can cover that, but it's the "kit" lens. On a tripod, it will be better of course! With the paper background, do I still need to use masking? I will also look at tutorials... Glad to know about "matches" viewer. This should look great when done right

[Andrea Alessi]

Hi!

I had a look at your dataset!

The issue here is mostly noise. You shot with Iso-1600 - if you zoom in and see your images at 100%, you'll see how noisy your images are. That's a problem cause that will cause a lot of false matches. You can use the Tools -> Workspace -> Matches viewer tool to see how many matches are made fro image to image. You'll see that there are very few, and this makes difficult for zephyr placing all cameras.

When images are not sharp, you'll have to use more aggressive settings, such as Deep rather than default.

However, i recommend using better photos if possible in order to get the best results with 3DF Zephyr - since you're shooting with turntable, two tips

1. use a tripod when possible - rather than shooting 1/60s f/14 Iso-1600, try using a different exposure settings - when shooting indoors for example i generally try using 1 or 2 seconds, Iso-100 and even f/22 for a better depth of field. It really depends on your lighting - if you're using long exposure, make sure to use a remote (or the countdown) so that you pressing the camera doesn't move it.

2. use masking to remove the background (you can do it completetly automatically with 3DF Masquerade) - when background and foreground are no longer coherent, zephyr will get confused as it doesn't know what you want to reconstruct (the wall? the statue?)

If you have a few minutes, I suggest you take a look at these tutorials:

photography for photogrammetry : https://www.youtube.com/watch?v=E06k...ature=youtu.be

masking for turntable scenarios : https://www.3dflow.net/technology/do...for-turntable/

As for the amount of images / overlap, it looks good - you just need sharper photos

Cheers,
Andrea