StereoScan performing much better than PhotoScan

[datapolo]

Hi,
I have been struggling over the weekend to get a decent model from 3 photographs in the demo of PhotoScan. The model tends to have waves and lumps in it that are not there in real life. I got it somewhere near acceptable but not what I expect. As a test I put two of the pictures in StereoScan and it produced a beautiful model that is what I expected. I then went back into PhotoScan and tried the same to photos rather than the 3. After doing the alignment I then went on to try to Build Geometry... and after a second or so it told me that it can't build the model!

I find this very surprising! Is there a simple explanation for this and what are the settings in PhotoScan that will mimic StereoScan!

Mike

[Dmitry Semyonov]

Hello Mike,

That seems strange as PhotoScan results should be comparable to StereoScan on a pair of images.

Can you please provide the photos you are processing to demonstrate the problem?

I've sent you an e-mail on this subject.

With best regards,
Dmitry Semyonov
AgiSoft LLC

[datapolo]

Hi Dmitry,
I sent a reply to your mail box about a week ago - I was wondering why you hadn't replied!

Anyway I got rid of the wave effect in PS by reducing the number of faces to between 4000 and 6000 and that gave a significant improvement but I still like the SS output better. Also for some pairs of photos PS misses out large parts of the model that SS includes and models well.

If there are basic settings in SS then it would be useful to know what they are if they can be used in PS to get a similar result.

Cheers,
Mike

[Dmitry Semyonov]

Hello Mike,

PhotoScan should produce the same results as StereoScan on a pair of images when Exact reconstruction method is used. Please also note that there is a "Decimate mesh" option in the StereoScan Preferences dialog that automatically reduces the number of faces in the final model when enabled.

If you still observe mismatch in the reconstruction results, please provide a pair of photos that demonstrate the problem.

P.S. We have replied your email on Wednesday, so I am just duplicating it here.

With best regards,
Dmitry Semyonov
AgiSoft LLC