It's great to see the forums here so full of life, active and people sharing cool ideas. Andreas has made some really great recommendations for tighter smaller set-ups using 35mm that I often reference my build clients to. Some of you guys are getting great results.
I hope the tutorial I posted on line is of some use. Quality is key. Especially to future proof for better 3D printers. At the moment they can't utilize all the detail and colour that a photogrammetry system can produce, but maybe some day!
Hopefully it's not like a button press. I hope it won't turn into a button press as Subunderground wishes. (Sorry Subunderground, don't take it personal) There are people like Lee and others who do this for living for good. I believe there must be difference between good work and shitty work.!
I agree with this to a point
The thing is for 4D motion scanning to ever be possible. 100,000 of frames! we need a push button solution. It has to be automated. I'm pretty sick of processing scans to a high level. It takes time and patience but at the moment is the only way.
I have tried very hard with Alexey to offer ideas for improvements and Agisoft LLC have been very patient but we have MANY hard problems still to solve.
My main BIG suggestions that could potentially automate the whole process are:
1) Better automated mask generation from background. Utilize the Region box or parallax to focus the Mask generation on the subject. Ignoring background elements. Pure, perfect, crisp masks are key! for 2.
2)
This is most important. Utilize Masks
during Geometry reconstruction AND
during hole fill. If we can have this, it will solve MANY problems. Hole fill and webbing being the biggest issue to solve but I believe this is a complex problem to solve.
3) Complex smoothing algorithm, something custom that can utilize the colour information and masks to refine silhouette but keep high frequency details..
4) Network, batch processing.
Just some ideas.