Next Version Update


#1

Hi there,

We are now in the third quarter off 2018.

  • I am wondering if there is going to be a version 2.xx of Sequoia?
  • Is there a roadmap?
  • Or is Sequoia very dead?

Thank you


#2

I’ll have to check here. We’re still placing items in the issue tracker, but I’m not sure on the roadmap at the moment.


#3

Any news regarding version 2.xxx of Sequoia?


#4

We’re not sharing roadmaps details at the moment as needs always change, but we’re continuing to support Sequoia for the foreseeable future. That includes bug fixes and new features.

I guess, what are you looking for in a Sequoia 2.0?


#5

I am glad you asked. Thank you.

Features:

  • Autodesk has released the SDK for their ReCap format RCP.
    Importing ReCap files in native its format directly into Sequoia with normals preserved would be a good thing.

  • Adding interpolation as a switchable function of the meshed so that logical holes are filled during the meshing process. This would likely fill the gaps in windows and occlusion shadows.

  • Photogrammetry

  • Importation of camera positions in FBX format from which to project textures frame and panoramic

  • Write manual and finish out the promised tutorial videos

Fixes:

  • Normals function is still in Beta and doesn’t work very well. Particularly around the scanner crop circle at tripod base.

  • Fix projection inaccuracies particularly with panos. The importation of externally solved camera would help here. Alternative if you provide solid photogrammetry solving then matching LiDAR to texture would be fully solvable in Sequoia without manual intervention.

  • Fix the normals based occlusion texture masking. The mask never seems to be fully opaque and now method for adjusting for that. Truly great idea though.

  • Finer control over UV generation on a per hacksaw tile basis with UDIM support.

  • support for 16 and 32 bit EXR images as texture source. Both frame and panorama

If I had to choose the highest priority features they would be Photogrammetry, solved camera importation, RCP format support and fixed projection mapping with refined occlusion masking.

Sequoia is a very solid product except it falls down on some critically important areas. Texture projection is a real fail in my opinion. So too is the Create Normals function. After all you can’t create single sided mesh without normals so what is the point?

I need to create fully textured very high resolution @32bit/16bit meshed models from point cloud.
When I say high resolution a pretty average pano for me weights in at over 340 Megapixels and there would from 10 to 30 of these per environment ideally.

At the moment I solve for the panos in one app then solve for the camera positions in another then register the LiDAR scans in ReCap. Recap or the scanner native software exports without normals so off to yet another app to recalculate the normals and align the LiDAR cloud with the Photogrammetry camera solved cloud. Once compete the LiDAR data is meshed and exported to yet another app for UVs then back again for texturing through reprojection from the solved panos and/or frames.

Totally crazy talk madness! However the process does hit the highest resolution photorealism.
With just a few changes and fixes Sequoia could alleviate either all or most of this pain.
While being fully supported pipeline wise through Deadline (Bliss).

Thank you for the opportunity to speak.
Trev


#6

Some great ideas there! The photogrammetry is quite the ask though. I’m not sure if there are libraries for this at the moment (please let me know if there are!), so cooking things up from scholarly articles is going to be some work.

I need to digest all of the asks here and split them into dev issues, but I’ll get to this when I can.


#7

Hi Edwin,

Yes photogrammetry is quite the ask. A possible library source http://www.regard3d.org

Alternatively I can use Photoscan Standard from Agisoft http://www.agisoft.com

Where I solve for the cameras and export as FBX with their point cloud as .e57. I wonder could ThinkBox licence their technology – as a plug-in? (Just a thought)

I was hoping to then import the FBX into Sequoia so that the texture re-projection nodes will be sorted into the correct position/rotation relative to each other.

Then use the Sequoia alignment methods to align the cameras as a group using their imported point cloud to the LiDAR sourced point cloud. With a choice to fine tune the cameras individually if necessary. Making in my mind the importing of camera positions through FBX a far greater priority over photogrammetry.

However having the cameras solved within Sequoia would be the superior approach.

Particularly if you need to combine UAV (drone) sourced photogrammetry and terrestrial LiDAR.

The issue here is finding a method for bringing in pre-solved camera positions to simplify texture re-projection.

I found when evaluating Sequoia that the camera position wouldn’t solve correctly using existing methods.

Causing quite the effort and many retries to achieve an acceptably accurate texture alignment.

One thing that pScan does over all other solutions I have looked at is that Photoscan is the only engine that can solve a group of nodal cameras as panoramic.

Meaning that it can solve from pre-stiched Lat Long panos (Professional version only) as single cameras and/or solve the individual nodal cameras as though they where a pano (Standard and Professional versions). It can do this wether the cameras make up a full pano or not. Put another way it can solve for any set of overlapping nodal cameras. This is an intensely useful feature.

Another big thing is native RCP (ReCap) support with normals. This would remove the need to post generate normals from ReCap sourced point clouds. Currently a big time waster. Autodesk have or are in the process of releasing the SDK for RCP.

I guess the above is the long winded way of saying please add native import of RCP files with normals intact and FBX support for importing and aligning solved camera positions to assist in texture re-projection.
Imported as either single camera Lat Long panos or multi camera nodal positions.

Thank you.
Trev


#8

Heh. Fair enough! I’ve converted some of the asks into issues on our side and forwarded this thread to one of the developers for the topics I don’t know as well.


#9

That is mighty good of you there Edwin. I look forward to seeing the outcome.
If there are any questions I’d be happy to respond.

Trev.