Dear @Vlad【ssh4】, as explained before, this is not the place for technical discussions, please use the forum or the github issue/enhancements pages for such comments.
@openmvs thanks for the feedback! When I say "pipeline for photogrammetry on sketchfab" I mean, what if we could let people upload a dataset of pictures and do the sfm/mesh extraction/optimization/texturing for them on sketchfab or in a "lab".
From what I've seen, you currently either have to glue software to create your own pipeline, or buy expensive licences. This is too cumbersome for most people. I know there is 123D catch but last time I tried the experience was not great. I'll have to test a bit and post on the forum if I have more precise questions then!
Yes, the steps are those in the wiki, with the mention that DenseReconstruction step was not done.
Estimate for this dataset on a normal computer is 20min for the complete OpenMVS pipeline.
Not sure I understand that you mean by "pipeline for photogrammetry on sketchfab". Could you please elaborate on [OpenMVS Group](https://groups.google.com/forum/#!forum/openmvs)
The result is very nice indeed! How come the mesh topologies are different on both geometries?
Also, I know this is asking a lot, but would you mind sharing the whole workflow you used? Is this the same as the one described in github.com/cdcseacave/openMVS/wiki/Usage...
I've been wondering for a while if we could easily put a a fully automatic and reasonable (CPU/RAM-wise) pipeline for photogrammetry on sketchfab but never found the time to test things yet. You would probably have very valuable input.
Could you give some estimates regarding time and ressources for this reconstruction?
17 comments
@openmvs nicee! do you have discord or an email I could contact you at?
@Victor.Colon.Cruz sure, what do u have in mind?
amazing work, love the gargoyle! btw do you accept any concept art ideas by any chance?
Nice work, made a 3D printable version here: thingiverse.com/thing:2439613
Omg, very good.
Great model!
@openmvs 了解しました！
Dear @Vlad【ssh4】, as explained before, this is not the place for technical discussions, please use the forum or the github issue/enhancements pages for such comments.
@mrchlblng
this is common log from OpenMVS (RefineMesh)
17:27:56 [App ] Scene loaded (375ms):
106 images (106 calibrated) with a total of 1626.02 MPixels (15.34 MPixels/image)
****
19:36:52 [App ] Mesh saved: 206690 vertices, 412926 faces (150ms)
19:36:52 [App ] MEMORYINFO: {
19:36:52 [App ] PageFaultCount 980168033
19:36:52 [App ] PeakWorkingSetSize 28.55GB
19:36:52 [App ] WorkingSetSize 3.73GB
19:36:52 [App ] QuotaPeakPagedPoolUsage 16.34MB
19:36:52 [App ] QuotaPagedPoolUsage 16.34MB
19:36:52 [App ] QuotaPeakNonPagedPoolUsage 3.47MB
19:36:52 [App ] QuotaNonPagedPoolUsage 3.39MB
19:36:52 [App ] PagefileUsage 5.48GB
19:36:52 [App ] PeakPagefileUsage 57.75GB
19:36:52 [App ] } ENDINFO
Memento/Remake on highest quality use not less than 16-24Gb of RAM.
Only ContextCapture and CapturingReality have really good memory and speed optimisation, and usually do not need more than 8-16Gb of RAM.
But i will glad if someone optimize OpenMVS for similar requirements :)
@mrchlblng I'd gladly contribute to such a project, but you have to detail it a bit...
@ssh4 thanks for the link!
@openmvs thanks for the feedback! When I say "pipeline for photogrammetry on sketchfab" I mean, what if we could let people upload a dataset of pictures and do the sfm/mesh extraction/optimization/texturing for them on sketchfab or in a "lab".
From what I've seen, you currently either have to glue software to create your own pipeline, or buy expensive licences. This is too cumbersome for most people. I know there is 123D catch but last time I tried the experience was not great. I'll have to test a bit and post on the forum if I have more precise questions then!
@openmvs people's still think that dense cloud is final step in defining topology, because 99% use Agisoft Photoscan :)
@mrchlblng there is basic tutorial:
From VisualSFV to OpenMVS
groups.google.com/forum/m/#!topic/vsfm/V...
The mesh topology changes during mesh refinement.
Yes, the steps are those in the wiki, with the mention that DenseReconstruction step was not done.
Estimate for this dataset on a normal computer is 20min for the complete OpenMVS pipeline.
Not sure I understand that you mean by "pipeline for photogrammetry on sketchfab". Could you please elaborate on [OpenMVS Group](https://groups.google.com/forum/#!forum/openmvs)
The result is very nice indeed! How come the mesh topologies are different on both geometries?
Also, I know this is asking a lot, but would you mind sharing the whole workflow you used? Is this the same as the one described in github.com/cdcseacave/openMVS/wiki/Usage...
I've been wondering for a while if we could easily put a a fully automatic and reasonable (CPU/RAM-wise) pipeline for photogrammetry on sketchfab but never found the time to test things yet. You would probably have very valuable input.
Could you give some estimates regarding time and ressources for this reconstruction?
Thx! Yes, all default settings, but the dense point-cloud step was skipped.
Nice result!
Is this default settings?