dizzux thanks for your vote. Photosculpt should improve concerning hard edges. Doing it manually is maybe the solution. I intend to do a prototype first. Who else is interrested?
this is now done for the obj export. See it in menu/?/laboratory in pay version of v1.03
Comments are welcomed.
Right now It’s for obj format only. Do you need it for other formats?
As of today I'm lacking feed-back from the 3D coat community. I don't know if everyone is happy. It's still quite a prototype status now.
My worry nr 1 is that I tried to sculpt the closed model with other sculpt software and (IMHO) it's not good as it makes very long trias close to very short ones.
So maybe it's only good for 3D coat?
2nd worry is that it does not work with the alpha selection tool. (that could be cool to mix both solutions together! But programmatically this is an entirely different story though)
Guys what do you think I should do?
interesting link Punkd.
No loss of performance? That could well be a starting point for Photosculpt
Hi, I'd love to but it's not possible right now.
Migrating to Mac is not impossible but it's difficult.
Let me explain what's the roadblock:
As an independent developper and self editor of Photosculpt. I don't have that "firepower" today, be it knowledge-wise, hardware-wise, dedicated staff-wise, time-wise or investissment-wise. Photosculpt is new (created in 2010) and is starting small like most/all start-ups.
Photosculpt is a pro tool witch is appreciated and also increasing in audience so at some point it may be feasible to invest in the porting to Mac. But being realistic, I'm sorry it may take months or even years.
Many pro softwares are not running on Mac too for the same reasons I believe.
Alternatively, if a contributor or editor is interested to take the challenge and help me port photosculpt to Mac, I'm ready to discuss. First thing would be to clarify the roadmap.
Thanks for dropping a word along with your vote.
Who else want a Mac version?
The more votes the better
I see exactly what you mean and I have good news. I've improved Photosculpt recently (internal revision v1.06, currently being tested).
it uses Exif focal length. And you can specify the focal lenght manually too (you can use it if you happen to have the real calibrated value instead of the constructor's value witch is only estimated +-5%)
Also: Photosculpt v1.06 can now compute (solve) the focal length from 2 photos for you. No need to imput key points so far, it's runs automatic.
It additionaly computes the lens barrel or pincushion deformation if any. and camera 6dof positions in 3D space.
The target is to help Photosculpt to be more "metric"
This version is not released yet but do you want to try it? You may help with your experience.
Just an update: My first trial with the mean of the 2 models is no good. Yes the models are combined but the artifacts are combined also = 2 times more artifacts. To improve the 3d model I need to do it the hard way: triangulate in 3 dimensions. For this now I think that I need a good 3D equation solver. So I builded one and started to tune it but it is not precise enough (yet). Stay tuned for more? If someone went through this already, help would be appreciated. :-)
My initial idea was to propose to sculpt from Left + Front + Right photos.
I will need to go thru prototypes and rework a lot of things but I'm willing to try this.
The most difficult thing will be to settle with what from the left or the right photo will have to take the lead for sculpting the final model. It can lead to monstruosities ;-)
One user (Dave) proposed to make a mean of the 2. That's simple enough to be tested first and after that we'll see. Any idea?
If I do prototypes they will be in the menu/?/under dev section as usual
Cool. I actually found pftrack interesting for solving camera position. Can it do depth maps to?
hello, thanks for posting this idea digipix!
Users usually vote for ideas and don't post new ones so this is cool that you took time really think about that.
warning you're mixing those are 2 entirely separate fuctions:
>function 1. you have a sequence of stereo pair images (a 3d film) as input and want a depth map of each pair (notice with photosculpt textures today you'll have to choose either for the left image or the right image or even both)
I should be able to make photosculpt work as batch and process images in sequence. That's not the main issue. The big issue is that you'll have a mix of good and bad images. Remember that photosculpt need specific shooting conditions and specific subjects (see tutorial 2). We may try this one day when batch will be implemented but I fear this is not going to be very good.
Why not try it entirely manually first? Who has a 3d film as image files and want to try
>function 2 you have a regular 2D film and want to extract the 3D model (notice: not the depth map) of what's been filmed, (supposing the subject is not moving)
I saw 2 videos: the church cube model modeling(all automatic) and the SUV modeling (all manual with markers) Wow this is all cool but it's really far away for photosculpt to do this soon. let's think. There is a major issue with data structure (pst stores in memory a depth map only) Any idea?
you are right, this can accelerate sculpting time greately.
A the moment I have not investigated this as Multithread was a priority for me. This comes as second. But now multithread is almost done, it's actually in testing phase now.
Who else want GPU acceleration?
Prototype is done and tests are ongoing, should reasonably be part of v1.07 and above
Hi Chris, many thanks for your explainations per email too. I like the idea. I'll think about it. I also think it can be done without a major rework of current functions.
I fear this file format to be a lot heavier than obj format for big meshes as color is added in ascii text. But as you said, it may be handy for merging meshes together.
Thanks for your suggestion. Who else is interested?
agreed, Colada would be nice on the feature list
For this I need a c++ code or visual c++ library to do this, I haven't found any nor tested any. Any advice somebody? This would really help me decide to investigate.
This is now implemented in v1.03. This is actually very experimental as some blur is needed for the vision algorithm mathematical stability.
You can play freely with this parameter now but it might now work as you wish. See my blog for a sample
yes I think like you. It will also work well on some subjects (textured with sharp edges)
Note that some minimal "blurring effect" is needed for the mathematical stability of the algorithms. It means that in some cases it might not work as you want. This is why I think a prototype is a good thing.
Hi Martin, that's a good idea. I'll think about it.
To implement that, I may do a prototype first (for anyone to test), assess quality, then connect it to the interface in a second step.
I asked Andrew and he told me he's thinking about it. Thanks!
I don't know how to do this. Please guys help me, send me examples or algorithms on how you want it done. I'm in the dark.
Improved in v1.03
More improvement and functionality planned for v1.06
Improved Edit Sculpt will be provided on release v1.02
But it won't be a full feature yet and still in the "Menu / ? / under dev".
Thanks for voting! ;-)
Prototype phase started for a batch tool. (maybe integrated in v1.07)
Command line functionality is planned for v1.06
Hi Dave, yes that would help me to see "Exactly" what you mean and not miss something important. If you get the time I would appreciate because that's what I need now: a visual for an interface concept.
Oh yes it does.
"Droplet" program: I love programs that just work without asking questions.
--> What about droping a set of files or an entire directory?
Command line: The good thing about command lines is that it's reasonably simple to program. But I never use them. As a user I want cool interfaces. That will only be a bonus for programmer and scripters.
You're the first to ask for command lines ;-) Let me add that as a new function if you haven't already. It's not quite the same as having batch processing. It might get some new feed back as solo item. thks for writing!
I need help: can someone design an interface, even on a piece of paper. Do you want it as a separate product or integrated? What should it contain ?
It’s done. I’m testing this now with (fabulous) results on my machine, cuting up time by -40% or so. This improvement will be probably included for free in the next subversion (1.05?)
Hi note that the sculpting is done by one core. The interface and the 3d model display that deals with Directx is done on another core.
So today you have a "multicore with 2 cores only" software ;-)
Here I mean sculpting with more cores. Splitting the sculpting job and only this one. So it goes faster.
This mean reworking a lot of functions as it's not an obvious task to pilot.
On my side I really thing this has to be done. But it will take time. Please vote if you want to prioritise that?
Added in v1.06
Greg proposed this on the thread about batch processing
Photosculpt.exe -front "c:\image1.jpg,image2.jpg" -right "c:\image3.jpg,image4.jpg" -output "c:\output_folder\output_file" -obj -depth -normal -spec -diff -occ -lod 0 -minimum
when Photosculpt is ran with the argument "-minimum", only a progress bar will show (that can be minimized), saving some memory by not showing the geometry and images on screen.
some progress was made
1. camera positions solver in v1.06
2. focal length correction in v1.06
Contact me at firstname.lastname@example.org if you want to contribute by testing v1.06 ?
We have to plan. We need to go thru it step by step. So what's the next step?
May I propose doing: left + center + right photos as first sub-objective?
This is extremely difficult to do. This may take a long time to make it work correctly with lot of R&D and investment. Any help is welcomed on that subject?