Page 1 of 1

Testing Photoscan

Posted: 04 May 2017, 09:11
by Curiousjeff
I have poor memories of trying Photogrammetry before using David (HP 3D).

I have now given it a try with Photoscan and my famous plastic gorilla.

I modified Walter's turntable to add automatic continuous rotation and a IR emitter to trigger my Canon camera.

Computing time is very long, but actual scanning time is short.

30 photos per 360 rotation

4 different positions of camera (from below, front 1, front 2, and from above).

Total photos: 120

I am impressed. I tried to load the same photos in Autodesk Remake but the result was horrible.

Jeff

Re: Testing Photoscan

Posted: 04 May 2017, 12:59
by Micr0
That's not bad at all for photogrammetry. Can you post a picture of the best scan you got with SL for a side by side comparison?

Re: Testing Photoscan

Posted: 04 May 2017, 16:50
by avogra
wow, im surprised! i never tried photogrammetry and had the opinion, that it needs texture to find height information. so only where texture is detailed, you can get a detailed surface. quite wrong apparently. is there a trial version of photoscan?

Re: Testing Photoscan

Posted: 04 May 2017, 18:34
by mading
For me it's unbelievable!
Really a nice results.

I built a turntable weeks ago for a friend.
Now I'm thinking about making one for myself!
I guess David 5 does not support DIY turntables.
Is it still possible to automate it using a time delay?

Re: Testing Photoscan

Posted: 04 May 2017, 19:22
by Curiousjeff
@ Micr0

After a crash, I lost some of my trial scans some time ago. Here is a photo from the old forum. I will make a test with David soon.

"From left to right

Canon Live view
Canon image grabber 21 patterns
Nikon fullHD + Blackmagic Intensity Pro 4K HDMI capture card
GH4 4k + Inogeni 4K HDMI capture card"

I am sure that David is more precise and you get the right scale out of the box. But depending on the purpose, this is a very quick solution (in terms of man power). I could take the photos of 10 items in less then two hours. Nearly all the rest would be automated. And the texture is great since it is build from high resolution photos under controled lighting (I did not post the texture here).

Anywere where precision is not paramount, this is a very interesting alternative.

@ avogra

Yes, there is a trial version of Photoscan. The purchase price of the basic version is $ 180.-
I need to confirm but I think it has everything needed for this kind of scan.
I don't thing you need any special kind of surface, as long as it is not shiny or transparent.
Many marble statues on Sketchfab were "Scanned" with Photoscan.

For example here:

https://sketchfab.com/egiptologo91

@ Madding

With Walter's turntable and Autoit, you can automate David. Maybe the code of the Arduino needs to be modified (?)

Or not, if David 5 can scan automatically without feedback from the turntable, then it would be simple to change Walter's code on a time base.

Jeff

Re: Testing Photoscan

Posted: 05 May 2017, 09:36
by Curiousjeff
It's actually scary, because the model from photoscan has no scale.

I did a very rough up scale using meshlab. Highly inaccurate.

Nevertheless, David was able to overlap them with "free align" (ok, the dimensions are not perfectly identical.)

Yellow: Photoscan
Purple: David with GH4 (full hd) and fusion [not a 100% sure, because this was done some time ago)

I added one from photo from the set given to Photoscan. Maybe I could improve the model with better exposed images.

And yes, I know that my gorilla is bizarre, The front was badly spray painted.

Jeffe.

Re: Testing Photoscan

Posted: 05 May 2017, 12:49
by mading
Thanks for sharing Jeffe!
I'm looking forward to do the same on my violin scroll and to post results.

m

Re: Testing Photoscan

Posted: 10 May 2017, 16:45
by Curiousjeff
Another test with a large item.

A wood mask.

I think it is a good alternative to David, when the main purpose is presentation on the web or things like that.

I really enjoy the time saving for the operator once the workflow is mastered.

The mask against the grey background is the 3D model with texture.

The mask against the black background is of the photos from the set of 150.

I did not use the highest setting, because of the processing time.

There are some surface errors (the pits on the nose). I am not sure if this is because is the shiny surface or something in my workflow.

Jeff

Re: Testing Photoscan

Posted: 10 May 2017, 17:27
by Curiousjeff
You can view it here:

https://skfb.ly/6qB8Z