First off this is mix of suggestions and questions..so not really sure if it should be here or the general forum
I've been using Photoscan for quite a while now,I'm very happy with the mesh it builds...but i have a few problems with the texturing side of things.
I'm working with a fixed camera rig, only 3 cams at the moment but will be expanding it.
I am photographing faces/heads..the person sits inside a dome of controlled LED lights that i have built,the lights are programmed to run through a sequence whilst the cams take a shot at every step of this sequence.
The setup is similar to Debevecs 'Lightstage'...the point of the 'special' lighting is to achieve a normal map that can add a high level of fine detail to the mesh that i get from photoscan.
I use the first shot(GI-all lights on) from each camera in photoscan to build the mesh.
My first issue is that if i use photoscan to build the UV's and textures then the blending is really not very good...it gives a lot of visible smears and almost semi transparent 'ghosting'.
What is the method used for the blending?...i'm interested in this because i have built a rig/node setup in cinema 4D that blends photos that are camera mapped to a head mesh and it seems to work a bit better than the photoscan blending even on a much lower res mesh.
At the moment my only alternative in photoscan is to use the 'single photo' option to make the texture from each camera then export to photoshop to clean up.
To aid with this method it would be great if a layered file could be exported with masks for the image from each camera.
Due to my 'special light rig' i also have another problem with the texturing in photoscan.
Once i have built my mesh using the GI photo from each camera..i want to be able to swap the GI image out and replace with another image lit differently but taken from the same camera..to generate other texture maps on the same mesh
The only way i can find to do this is by renaming my photo..which is a very clumsy and confusing method...is there an alternative that does not involve rebuilding the mesh?
I know that its maybe a very specialist request, but it would be great to make this image swapping easier..it can't be that complicated can it?
General questions and a bug? :
When exporting model and cameras from photoscan using collada....if i tick the 'separate chunks' in the export options, then when I open the file in cinema 4D i get a model...if i don't tick this box, the file is opened by cinema 4D but contains nothing...but i don't want 'separate chunks' just one mesh.
There seems to be some strange random rotation and scaling issues with the exported models+cams
everything is often rotated by 90? and the axis is never in the middle of mesh...this can be a real pain,can anything be done about this?
Which parts of the build process are multi-threaded and whats the roadmap on this,will there be any changes?
Sorry if that sounds like a long list of complaints,but i do really love photoscan..the level of detail i'm getting using just a few photos is amazing.
Cheers,
mala